DeSola_Vinicio_Assignment02.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "IPython.OutputArea.prototype._should_scroll = function(lines) {\n",
       "    return false;\n",
       "}"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%javascript\n",
    "IPython.OutputArea.prototype._should_scroll = function(lines) {\n",
    "    return false;\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Project 2: Topic Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this project, you'll work with text data from newsgroup postings on a variety of topics. You'll train classifiers to distinguish between the topics based on the text of the posts. Whereas with digit classification, the input is relatively dense: a 28x28 matrix of pixels, many of which are non-zero, here we'll represent each document with a \"bag-of-words\" model. As you'll see, this makes the feature representation quite sparse -- only a few words of the total vocabulary are active in any given document. The bag-of-words assumption here is that the label depends only on the words; their order is not important.\n",
    "\n",
    "The SK-learn documentation on feature extraction will prove useful:\n",
    "http://scikit-learn.org/stable/modules/feature_extraction.html\n",
    "\n",
    "Each problem can be addressed succinctly with the included packages -- please don't add any more. Grading will be based on writing clean, commented code, along with a few short answers.\n",
    "\n",
    "As always, you're welcome to work on the project in groups and discuss ideas on the course wall, but please prepare your own write-up and write your own code."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to\n",
      "[nltk_data]     C:\\Users\\penpe\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Unzipping tokenizers\\punkt.zip.\n"
     ]
    }
   ],
   "source": [
    "# This tells matplotlib not to try opening a new window for each plot.\n",
    "%matplotlib inline\n",
    "\n",
    "\n",
    "# General libraries.\n",
    "import re\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import pandas as pd\n",
    "from IPython.display import display, HTML, Markdown\n",
    "from scipy.sparse import csr\n",
    "from scipy.sparse import csr_matrix\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.tokenize import word_tokenize \n",
    "nltk.download('punkt')\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "# SK-learn libraries for learning.\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# SK-learn libraries for evaluation.\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn import metrics\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "# SK-learn library for importing the newsgroup data.\n",
    "from sklearn.datasets import fetch_20newsgroups\n",
    "\n",
    "# SK-learn libraries for feature extraction from text.\n",
    "from sklearn.feature_extraction.text import *\n",
    "\n",
    "# Display centered pandas tables\n",
    "def pd_centered(df):\n",
    "    return df.style.set_table_styles([\n",
    "        {\"selector\": \"th\", \"props\": [(\"text-align\", \"center\")]},\n",
    "        {\"selector\": \"td\", \"props\": [(\"text-align\", \"center\")]}])\n",
    "# Print in bold\n",
    "def printmd(string):\n",
    "    display(Markdown(string))\n",
    "\n",
    "# Plot formats\n",
    "from cycler import cycler\n",
    "import matplotlib.patches as patches\n",
    "import matplotlib.mlab as mlab\n",
    "import matplotlib.dates\n",
    "plt.style.use('ggplot')\n",
    "mpl.rcParams['lines.linewidth'] = 1.5\n",
    "mpl.rcParams['lines.color'] = 'blue'\n",
    "mpl.rcParams['axes.prop_cycle'] = cycler('color', ['#30a2da', '#e5ae38', '#fc4f30', '#6d904f', '#8b8b8b'])\n",
    "mpl.rcParams['legend.fancybox'] = True\n",
    "mpl.rcParams['legend.fontsize'] = 14\n",
    "mpl.rcParams['axes.facecolor'] = '#f0f0f0'\n",
    "mpl.rcParams['axes.labelsize'] = 15\n",
    "mpl.rcParams['axes.axisbelow'] = True\n",
    "mpl.rcParams['axes.linewidth'] = 1.2\n",
    "mpl.rcParams['axes.labelpad'] = 0.0\n",
    "mpl.rcParams['axes.xmargin'] = 0.05  # x margin.  See `axes.Axes.margins`\n",
    "mpl.rcParams['axes.ymargin'] = 0.05  # y margin See `axes.Axes.margins`\n",
    "mpl.rcParams['xtick.labelsize'] = 14\n",
    "mpl.rcParams['ytick.labelsize'] = 14\n",
    "mpl.rcParams['figure.subplot.left'] = 0.08\n",
    "mpl.rcParams['figure.subplot.right'] = 0.95\n",
    "mpl.rcParams['figure.subplot.bottom'] = 0.07\n",
    "\n",
    "# figure configuration\n",
    "fsize = (10,7.5) # figure size\n",
    "tsize = 18 # title font size\n",
    "lsize = 16 # legend font size\n",
    "csize = 14 # comment font size\n",
    "grid = True # grid\n",
    "\n",
    "# Set the randomizer seed so results are the same each time ... Reproducibility\n",
    "np.random.seed(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the data, stripping out metadata so that we learn classifiers that only use textual features. By default, newsgroups data is split into train and test sets. We further split the test so we have a dev set. Note that we specify 4 categories to use for this project. If you remove the categories argument from the fetch function, you'll get all 20 categories."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training label shape: (2034,)\n",
      "test label shape: (677,)\n",
      "dev label shape: (676,)\n",
      "labels names: ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n"
     ]
    }
   ],
   "source": [
    "categories = ['alt.atheism', 'talk.religion.misc', 'comp.graphics', 'sci.space']\n",
    "newsgroups_train = fetch_20newsgroups(subset='train',\n",
    "                                      remove=('headers', 'footers', 'quotes'),\n",
    "                                      categories=categories)\n",
    "newsgroups_test = fetch_20newsgroups(subset='test',\n",
    "                                     remove=('headers', 'footers', 'quotes'),\n",
    "                                     categories=categories)\n",
    "\n",
    "num_test = len(newsgroups_test.target)\n",
    "test_data, test_labels = newsgroups_test.data[int(num_test/2):], newsgroups_test.target[int(num_test/2):]\n",
    "dev_data, dev_labels = newsgroups_test.data[:int(num_test/2)], newsgroups_test.target[:int(num_test/2)]\n",
    "train_data, train_labels = newsgroups_train.data, newsgroups_train.target\n",
    "\n",
    "print('training label shape:', train_labels.shape)\n",
    "print('test label shape:', test_labels.shape)\n",
    "print('dev label shape:', dev_labels.shape)\n",
    "print('labels names:', newsgroups_train.target_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (1) \n",
    "\n",
    "For each of the first 5 training examples, print the text of the message along with the label."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "**Number of training examples to print: 5**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of training example 1**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*comp.graphics*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 1**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hi,\n",
      "\n",
      "I've noticed that if you only save a model (with all your mapping planes\n",
      "positioned carefully) to a .3DS file that when you reload it after restarting\n",
      "3DS, they are given a default position and orientation.  But if you save\n",
      "to a .PRJ file their positions/orientation are preserved.  Does anyone\n",
      "know why this information is not stored in the .3DS file?  Nothing is\n",
      "explicitly said in the manual about saving texture rules in the .PRJ file. \n",
      "I'd like to be able to read the texture rule information, does anyone have \n",
      "the format for the .PRJ file?\n",
      "\n",
      "Is the .CEL file format available from somewhere?\n",
      "\n",
      "Rych\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of training example 2**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*talk.religion.misc*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 2**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "Seems to be, barring evidence to the contrary, that Koresh was simply\n",
      "another deranged fanatic who thought it neccessary to take a whole bunch of\n",
      "folks with him, children and all, to satisfy his delusional mania. Jim\n",
      "Jones, circa 1993.\n",
      "\n",
      "\n",
      "Nope - fruitcakes like Koresh have been demonstrating such evil corruption\n",
      "for centuries.\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of training example 3**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*sci.space*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 3**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " >In article <1993Apr19.020359.26996@sq.sq.com>, msb@sq.sq.com (Mark Brader) \n",
      "\n",
      "MB>                                                             So the\n",
      "MB> 1970 figure seems unlikely to actually be anything but a perijove.\n",
      "\n",
      "JG>Sorry, _perijoves_...I'm not used to talking this language.\n",
      "\n",
      "Couldn't we just say periapsis or apoapsis?\n",
      "\n",
      " \n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of training example 4**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*alt.atheism*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 4**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I have a request for those who would like to see Charley Wingate\n",
      "respond to the \"Charley Challenges\" (and judging from my e-mail, there\n",
      "appear to be quite a few of you.)  \n",
      "\n",
      "It is clear that Mr. Wingate intends to continue to post tangential or\n",
      "unrelated articles while ingoring the Challenges themselves.  Between\n",
      "the last two re-postings of the Challenges, I noted perhaps a dozen or\n",
      "more posts by Mr. Wingate, none of which answered a single Challenge.  \n",
      "\n",
      "It seems unmistakable to me that Mr. Wingate hopes that the questions\n",
      "will just go away, and he is doing his level best to change the\n",
      "subject.  Given that this seems a rather common net.theist tactic, I\n",
      "would like to suggest that we impress upon him our desire for answers,\n",
      "in the following manner:\n",
      "\n",
      "1. Ignore any future articles by Mr. Wingate that do not address the\n",
      "Challenges, until he answers them or explictly announces that he\n",
      "refuses to do so.\n",
      "\n",
      "--or--\n",
      "\n",
      "2. If you must respond to one of his articles, include within it\n",
      "something similar to the following:\n",
      "\n",
      "    \"Please answer the questions posed to you in the Charley Challenges.\"\n",
      "\n",
      "Really, I'm not looking to humiliate anyone here, I just want some\n",
      "honest answers.  You wouldn't think that honesty would be too much to\n",
      "ask from a devout Christian, would you?  \n",
      "\n",
      "Nevermind, that was a rhetorical question.\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of training example 5**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*sci.space*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 5**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AW&ST  had a brief blurb on a Manned Lunar Exploration confernce\n",
      "May 7th  at Crystal City Virginia, under the auspices of AIAA.\n",
      "\n",
      "Does anyone know more about this?  How much, to attend????\n",
      "\n",
      "Anyone want to go?\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P1(num_examples=5):\n",
    "    \"\"\" \n",
    "    1st Problem. Print the text of the message along with the label\n",
    "    Args:\n",
    "        Number of examples we want to run. Current default value num_examples = 5 (number of training examples)\n",
    "    Returns:\n",
    "        Text of the num_examples along with the label\n",
    "    \"\"\"\n",
    "    printmd(\"**Number of training examples to print: %d**\"%num_examples)\n",
    "    for i in range(num_examples):\n",
    "        printmd('**-**'*100+'\\n')\n",
    "        printmd('**Label of training example %d**'%(i+1))\n",
    "        printmd('*%s*'%newsgroups_train.target_names[train_labels[i]]+'\\n')\n",
    "        printmd('**Text of the message in training example %d**'%(i+1))\n",
    "        print(\"\\n\".join(train_data[i].split(\"\\n\")))\n",
    "        printmd('**-**'*100+'\\n')\n",
    "    \n",
    "P1()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (2) \n",
    "Use CountVectorizer to turn the raw training text into feature vectors. You should use the fit_transform function, which makes 2 passes through the data: first it computes the vocabulary (\"fit\"), second it converts the raw text into feature vectors using the vocabulary (\"transform\").\n",
    "\n",
    "The vectorizer has a lot of options. To get familiar with some of them, write code to answer these questions:\n",
    "\n",
    "a. The output of the transform (also of fit_transform) is a sparse matrix: http://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.sparse.csr_matrix.html. What is the size of the vocabulary? What is the average number of non-zero features per example? What fraction of the entries in the matrix are non-zero? Hint: use \"nnz\" and \"shape\" attributes.\n",
    "\n",
    "b. What are the 0th and last feature strings (in alphabetical order)? Hint: use the vectorizer's get_feature_names function.\n",
    "\n",
    "c. Specify your own vocabulary with 4 words: [\"atheism\", \"graphics\", \"space\", \"religion\"]. Confirm the training vectors are appropriately shaped. Now what's the average number of non-zero features per example?\n",
    "\n",
    "d. Instead of extracting unigram word features, use \"analyzer\" and \"ngram_range\" to extract bigram and trigram character features. What size vocabulary does this yield?\n",
    "\n",
    "e. Use the \"min_df\" argument to prune words that appear in fewer than 10 documents. What size vocabulary does this yield?\n",
    "\n",
    "f. Using the standard CountVectorizer, what fraction of the words in the dev data are missing from the vocabulary? Hint: build a vocabulary for both train and dev and look at the size of the difference."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "### Part a"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Description of the sparse matrix**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Dimensions of the sparse matrix is (2034,26879)"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Number of examples is 2034 and the size of the vocabulary is 26879"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Average number of non-zero features per example is 96.706"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Fraction of non-zeroes in the matriz is 0.00360"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part b"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**The 0th and the last feature string**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "The 0th string is *00*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "The last string is *zyxel*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part c"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "** Implement own vocabulary = [\"atheism\", \"graphics\", \"space\", \"religion\"] **"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Check the dimensions of the sparse matrix (2034,4)"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Number of examples is 2034 and the size of the vocabulary is 4"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Average number of non-zero features per example is 0.268"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part d"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Implement a n-gram vocabulary using bigram and trigrams**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Check the dimensions of the sparse matrix (2034,35478)"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Number of examples is 2034 and the size of the vocabulary is 35478"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part e"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Implement a pruned vocabulary (prune words that appear in less than 10 documents)**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Check the dimensions of the sparse matrix (2034,3064)"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Number of examples is 2034 and the size of the vocabulary is 3064"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part f"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "** Fraction of words in the dev set not in the Standard Vocabulary**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Fraction of words that are in the development set, but missing from the Standard Dictionary are              24.788 %"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P2():\n",
    "    \"\"\" \n",
    "    2nd Problem. Turn the raw data into features vectors using several steps\n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment\n",
    "    Returns:\n",
    "        The function will return a report with six outputs: \n",
    "            a.- Output of transforms in a sparse matrix\n",
    "            b.- The 0th and Last feature string\n",
    "            c.- Vocabulary with 4 words\n",
    "            d.- Extraction of bigram and trigram character features\n",
    "            e.- Pruned vocabulary (prune words that appear in fewer than 10 documents)\n",
    "            f.- Compare the vocabulary from the dev set and the training set\n",
    "    \"\"\"\n",
    "    # Part a - Transform to vector of features\n",
    "    printmd('### Part a')\n",
    "    printmd('**Description of the sparse matrix**\\n')\n",
    "    count_vector = CountVectorizer()\n",
    "    train_data_count = count_vector.fit_transform(train_data)\n",
    "    printmd(\"Dimensions of the sparse matrix is (%d,%d)\"%train_data_count.shape)\n",
    "    printmd('Number of examples is %d and the size of the vocabulary is %d'%(train_data_count.shape[0],\\\n",
    "                                                                              train_data_count.shape[1]))\n",
    "    printmd('Average number of non-zero features per example is %.03f'%train_data_count.getnnz(axis=1).mean())\n",
    "    total_elements = train_data_count.shape[0]*train_data_count.shape[1]\n",
    "    printmd('Fraction of non-zeroes in the matriz is %.05f'%(train_data_count.nnz/total_elements))\n",
    "    \n",
    "    # Part b - 0th and Last feature\n",
    "    printmd('### Part b')\n",
    "    printmd('**The 0th and the last feature string**')\n",
    "    names = count_vector.get_feature_names()\n",
    "    printmd('The 0th string is *%s*'%names[0])\n",
    "    printmd('The last string is *%s*'%names[-1])\n",
    "    \n",
    "    # Part c - Own Vocabulary with 4 words\n",
    "    vocabulary = [\"atheism\", \"graphics\", \"space\", \"religion\"]\n",
    "    own_count_vector = CountVectorizer(vocabulary=vocabulary)\n",
    "    train_data_count_own = own_count_vector.fit_transform(train_data)\n",
    "    printmd('### Part c')\n",
    "    printmd('** Implement own vocabulary = [\"atheism\", \"graphics\", \"space\", \"religion\"] **')\n",
    "    printmd('Check the dimensions of the sparse matrix (%d,%d)'%train_data_count_own.shape)\n",
    "    printmd('Number of examples is %d and the size of the vocabulary is %d'%(train_data_count_own.shape[0],\\\n",
    "                                                                              train_data_count_own.shape[1]))\n",
    "    printmd('Average number of non-zero features per example is %.03f'%train_data_count_own.getnnz(axis=1).mean())\n",
    "    \n",
    "    # Part d - Extract Bigrams and Trigrams\n",
    "    ngram_count_vector = CountVectorizer(analyzer='char', ngram_range=(2,3))\n",
    "    train_data_count_ngram = ngram_count_vector.fit_transform(train_data)\n",
    "    printmd('### Part d')\n",
    "    printmd('**Implement a n-gram vocabulary using bigram and trigrams**')\n",
    "    printmd('Check the dimensions of the sparse matrix (%d,%d)'%train_data_count_ngram.shape)\n",
    "    printmd('Number of examples is %d and the size of the vocabulary is %d'%(train_data_count_ngram.shape[0],\\\n",
    "                                                                              train_data_count_ngram.shape[1]))\n",
    "    \n",
    "    # Part e - Prune Words that appear in less than 10 documents\n",
    "    prune_count_vector = CountVectorizer(min_df=10)\n",
    "    train_data_count_prune = prune_count_vector.fit_transform(train_data)\n",
    "    printmd('### Part e')\n",
    "    printmd('**Implement a pruned vocabulary (prune words that appear in less than 10 documents)**')\n",
    "    printmd('Check the dimensions of the sparse matrix (%d,%d)'%train_data_count_prune.shape)\n",
    "    printmd('Number of examples is %d and the size of the vocabulary is %d'%(train_data_count_prune.shape[0],\\\n",
    "                                                                              train_data_count_prune.shape[1]))    \n",
    "    \n",
    "    # Part f - Using Standard Vocabulary, check fraction missing\n",
    "    dev_count_vector = count_vector.fit_transform(dev_data)\n",
    "    names_dev = count_vector.get_feature_names()\n",
    "    printmd('### Part f')\n",
    "    printmd('** Fraction of words in the dev set not in the Standard Vocabulary**')\n",
    "    printmd('Fraction of words that are in the development set, but missing from the Standard Dictionary are \\\n",
    "             %.03f %%'%(100*(len(names_dev)-len(list(set(names).intersection(names_dev))))/len(names_dev)))\n",
    "    \n",
    "    \n",
    "    \n",
    "    \n",
    "P2()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (3) \n",
    "Use the default CountVectorizer options and report the f1 score (use metrics.f1_score) for a k nearest neighbors classifier; find the optimal value for k. Also fit a Multinomial Naive Bayes model and find the optimal value for alpha. Finally, fit a logistic regression model and find the optimal value for the regularization strength C using L2 regularization. A few questions:\n",
    "\n",
    "a. Why doesn't nearest neighbors work well for this problem?\n",
    "\n",
    "b. Any ideas why logistic regression doesn't work as well as Naive Bayes?\n",
    "\n",
    "c. Logistic regression estimates a weight vector for each class, which you can access with the coef\\_ attribute. Output the sum of the squared weight values for each class for each setting of the C parameter. Briefly explain the relationship between the sum and the value of C."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "### Part a - kNN Neighbors Classifier"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Optimal value for k is 150 with a F1 score of 45.08 %"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245cce02d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABDoAAAExCAYAAACd0sLdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nU5cAKW0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245afc40e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part b - Multinomial Naive Bayes"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Optimal value for alpha is 0.0100 with a F1 score of 82.94 %"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245bc869780>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABDoAAAE3CAYAAABLiyHAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245b8450668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part c - Logistic Regression with L2 Regularization"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Optimal value for C is 0.50 with a F1 score of 77.83 %"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245c8275898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABDEAAAExCAYAAABlMjkqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245b7749748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "Relationship between sum of square weights values of each class anc C: "
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=0.0001** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *0.01*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *0.01*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *0.01*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *0.01*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=0.1000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *27.13*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *24.65*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *27.46*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *23.03*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=0.2000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *49.75*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *42.75*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *49.32*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *42.67*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=0.5000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *102.62*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *83.08*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *99.04*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *88.98*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=0.8000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *143.58*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *113.58*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *136.68*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *124.99*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=1.0000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *167.12*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *130.87*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *157.95*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *145.73*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=1.5000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *216.60*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *167.48*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *202.84*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *189.77*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=2.0000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *257.66*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *197.88*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *239.93*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *226.61*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=5.0000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *422.76*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *322.90*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *389.63*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *377.85*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=10.0000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *586.41*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *447.94*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *539.09*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *530.68*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "- For **C=15.0000** the sum of square weights values of each class is:"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 1\\. *alt.atheism* - Sum of Squared Weights: *698.36*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 2\\. *comp.graphics* - Sum of Squared Weights: *536.29*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 3\\. *sci.space* - Sum of Squared Weights: *643.62*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "   ... 4\\. *talk.religion.misc* - Sum of Squared Weights: *637.90*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P3():\n",
    "    \"\"\" \n",
    "    3rd Problem. Using the default features vector, train a kNN, a Multinomial Naive Bayes, and a Logistic Regression\n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment\n",
    "    Returns:\n",
    "        The function will return a report with three outputs: \n",
    "            a.- F1 Score for the kNN classifier, after using GridSearch\n",
    "            b.- F1 Score for the Multinomial Naive Bayes Classifier, after using GridSearch\n",
    "            c.- F1 Score for Logistic Regression Classifier using L2 Regularization, using GridSearch\n",
    "    \"\"\"\n",
    "    classes = ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n",
    "    mpl.rcParams['figure.figsize'] = 16,4\n",
    "    fig = plt.figure(figsize=(16,4))\n",
    "    \n",
    "    # Part a - kNN Classifier\n",
    "    knn_clf = Pipeline([\n",
    "            ('vect', CountVectorizer()),\n",
    "            ('clf', KNeighborsClassifier(n_jobs=-1)),\n",
    "    ])\n",
    "    k = [1,2,5,10,15,20,25,30,40,50,60,70,80,90,100,150,200]\n",
    "    param_grid = {\n",
    "        'clf__n_neighbors': k}\n",
    "    best_knn_clf = GridSearchCV(knn_clf,param_grid,cv = 5,scoring='f1_micro')\n",
    "    best_knn_clf.fit(train_data, train_labels)\n",
    "    results = pd.DataFrame(best_knn_clf.cv_results_,index = k)\n",
    "    printmd('### Part a - kNN Neighbors Classifier')\n",
    "    printmd('Optimal value for k is %d with a F1 score of %.02f %%'%(best_knn_clf.best_params_['clf__n_neighbors'],\\\n",
    "                                                                 100*best_knn_clf.best_score_))\n",
    "    ax = results.plot('param_clf__n_neighbors','mean_test_score', linestyle='-', marker='o', logx=True)\n",
    "    ax.set_title(\"Search of Hyperparameter k - Mean Test Score vs. Number of Neighbors\")\n",
    "    ax.set_xlabel('Log Number of Neighbors')\n",
    "    ax.set_ylabel('Mean Test Score of Grid Search')\n",
    "    plt.show()\n",
    "\n",
    "    \n",
    "    # Part b - Multinomial Naive Bayes\n",
    "    nb_multi_clf = Pipeline([\n",
    "            ('vect', CountVectorizer()),\n",
    "            ('clf', MultinomialNB()),\n",
    "    ])\n",
    "    alpha = [0.0, 0.0001, 0.001, 0.01, 0.1, 0.5, 1.0, 2.0, 10.0]\n",
    "    param_grid = {\n",
    "        'clf__alpha': alpha}\n",
    "    best_nb_multi_clf = GridSearchCV(nb_multi_clf,param_grid,cv = 5,scoring='f1_micro')\n",
    "    best_nb_multi_clf.fit(train_data, train_labels)\n",
    "    results = pd.DataFrame(best_nb_multi_clf.cv_results_,index = alpha)\n",
    "    printmd('### Part b - Multinomial Naive Bayes')\n",
    "    printmd('Optimal value for alpha is %.04f with a F1 score of %.02f %%'%(best_nb_multi_clf.best_params_['clf__alpha'],\\\n",
    "                                                                 100*best_nb_multi_clf.best_score_))\n",
    "    fig = plt.figure(figsize=(16,4))\n",
    "    ax = results.plot('param_clf__alpha','mean_test_score', linestyle='-', marker='o', logx=True)\n",
    "    ax.set_title(\"Search of Hyperparameter $\\\\alpha$ - Mean Test Score vs. $\\\\alpha$\")\n",
    "    ax.set_xlabel('Log $\\\\alpha$')\n",
    "    ax.set_ylabel('Mean Test Score of Grid Search')\n",
    "    plt.show()\n",
    "\n",
    "    \n",
    "    # Part c - Logistic Regression with L2 Regularization\n",
    "    log_reg_clf = Pipeline([\n",
    "            ('vect', CountVectorizer()),\n",
    "            ('clf', LogisticRegression(penalty='l2', n_jobs=-1)),\n",
    "    ])\n",
    "    C = [0.0001, 0.1, 0.2, 0.5, 0.8, 1, 1.5, 2.0, 5.0, 10.0, 15.0]\n",
    "    param_grid = {\n",
    "        'clf__C': C}\n",
    "    best_log_reg_clf = GridSearchCV(log_reg_clf,param_grid,cv = 5,scoring='f1_micro')\n",
    "    best_log_reg_clf.fit(train_data, train_labels)\n",
    "    results = pd.DataFrame(best_log_reg_clf.cv_results_,index = C)\n",
    "    printmd('### Part c - Logistic Regression with L2 Regularization')\n",
    "    printmd('Optimal value for C is %.02f with a F1 score of %.02f %%'%(best_log_reg_clf.best_params_['clf__C'],\\\n",
    "                                                                 100*best_log_reg_clf.best_score_))\n",
    "    fig = plt.figure(figsize=(16,4))\n",
    "    ax = results.plot('param_clf__C','mean_test_score', linestyle='-', marker='o', logx=True)\n",
    "    ax.set_title(\"Search of Hyperparameter C - Mean Test Score vs. C\")\n",
    "    ax.set_xlabel('Log C')\n",
    "    ax.set_ylabel('Mean Test Score of Grid Search')\n",
    "    plt.show()\n",
    "    printmd('Relationship between sum of square weights values of each class anc C: ')\n",
    "    for c in C:\n",
    "        log_reg_clf = Pipeline([\n",
    "            ('vect', CountVectorizer()),\n",
    "            ('clf', LogisticRegression(penalty='l2', C=c, n_jobs=-1)),\n",
    "        ])\n",
    "        log_reg_clf.fit(train_data, train_labels)\n",
    "        printmd('- For **C=%.04f** the sum of square weights values of each class is:'%c)\n",
    "        \n",
    "        for i, weight in enumerate(classes):\n",
    "            printmd('   ... %s\\. *%s* - Sum of Squared Weights: *%.02f*'%(i+1,weight,\\\n",
    "                                                                 np.sum(log_reg_clf.named_steps['clf'].coef_[i]**2)))\n",
    "        \n",
    "\n",
    "a = P3()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Answer:\n",
    "\n",
    "a. kNN doesn't work well in this problem because it needs to have a large dataset to be effective: in our case we are working with only 2034 examples. Therefore, it needs many neighbors to be effective, and is not finding out the real division of the hyperspace because of lack of training data. Also, the way distances work in the space is not well suited: how 'far' a word is from another is not well defined, because we are not normalizing the way we process the frequency count.\n",
    "\n",
    "b. Because the data fits a Naive Bayes Multinomial quite well: events are almost idenpendent, training data is small, and is possible the events comes from a non-linear data generative model, which means logistic regression will present some problems. However, the greater the amount of training data, logistic regression should perfom better\n",
    "\n",
    "c. Is obvious that the higher the value of C, the sum square weights of each class will increase. C is the inverse of the regularization parameter, therefore, the lower its value, the more penalized each weight are. When C increases, the weights increases with it. Because we are using a L2 regularization, this relationship is valid in the sum square space. Also, we can see that the the F1 score is flat around 0.1 to 0.5. The smaller the value of C, the weights become relatively smaller. For P4 I'll compromise and use a value of 0.25"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (4) \n",
    "Train a logistic regression model. Find the 5 features with the largest weights for each label -- 20 features in total. Create a table with 20 rows and 4 columns that shows the weight for each of these features for each of the labels. Create the table again with bigram features. Any surprising features in this table?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "### Part a - Logistic Regression - Word Vocabulary C = 0.25"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*Table of the top 5 features by weights per label*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<style  type=\"text/css\" >\n",
       "    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col0 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col1 {\n",
       "            background-color:  #565656;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col2 {\n",
       "            background-color:  #434343;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col3 {\n",
       "            background-color:  #4f4f4f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col0 {\n",
       "            background-color:  #fefefe;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col1 {\n",
       "            background-color:  #5e5e5e;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col2 {\n",
       "            background-color:  #454545;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col3 {\n",
       "            background-color:  #363636;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col0 {\n",
       "            background-color:  #fefefe;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col1 {\n",
       "            background-color:  #484848;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col2 {\n",
       "            background-color:  #3e3e3e;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col3 {\n",
       "            background-color:  #4d4d4d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col0 {\n",
       "            background-color:  #fefefe;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col1 {\n",
       "            background-color:  #3a3a3a;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col2 {\n",
       "            background-color:  #242424;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col3 {\n",
       "            background-color:  #7f7f7f;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col0 {\n",
       "            background-color:  #eeeeee;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col1 {\n",
       "            background-color:  #5e5e5e;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col2 {\n",
       "            background-color:  #444444;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col3 {\n",
       "            background-color:  #5d5d5d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col0 {\n",
       "            background-color:  #383838;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col1 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col2 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col3 {\n",
       "            background-color:  #2f2f2f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col0 {\n",
       "            background-color:  #535353;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col1 {\n",
       "            background-color:  #cecece;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col2 {\n",
       "            background-color:  #252525;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col3 {\n",
       "            background-color:  #515151;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col0 {\n",
       "            background-color:  #6b6b6b;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col1 {\n",
       "            background-color:  #cacaca;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col2 {\n",
       "            background-color:  #212121;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col3 {\n",
       "            background-color:  #404040;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col0 {\n",
       "            background-color:  #676767;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col1 {\n",
       "            background-color:  #bfbfbf;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col2 {\n",
       "            background-color:  #2d2d2d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col3 {\n",
       "            background-color:  #595959;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col0 {\n",
       "            background-color:  #939393;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col1 {\n",
       "            background-color:  #b9b9b9;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col2 {\n",
       "            background-color:  #2c2c2c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col3 {\n",
       "            background-color:  #4f4f4f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col0 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col1 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col2 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col3 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col0 {\n",
       "            background-color:  #616161;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col1 {\n",
       "            background-color:  #353535;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col2 {\n",
       "            background-color:  #b0b0b0;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col3 {\n",
       "            background-color:  #444444;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col0 {\n",
       "            background-color:  #535353;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col1 {\n",
       "            background-color:  #424242;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col2 {\n",
       "            background-color:  #a4a4a4;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col3 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col0 {\n",
       "            background-color:  #5f5f5f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col1 {\n",
       "            background-color:  #434343;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col2 {\n",
       "            background-color:  #9e9e9e;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col3 {\n",
       "            background-color:  #5d5d5d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col0 {\n",
       "            background-color:  #696969;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col1 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col2 {\n",
       "            background-color:  #979797;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col3 {\n",
       "            background-color:  #5f5f5f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col0 {\n",
       "            background-color:  #535353;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col1 {\n",
       "            background-color:  #4a4a4a;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col2 {\n",
       "            background-color:  #444444;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col3 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col0 {\n",
       "            background-color:  #414141;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col1 {\n",
       "            background-color:  #4d4d4d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col2 {\n",
       "            background-color:  #3c3c3c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col3 {\n",
       "            background-color:  #fcfcfc;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col0 {\n",
       "            background-color:  #5a5a5a;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col1 {\n",
       "            background-color:  #5f5f5f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col2 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col3 {\n",
       "            background-color:  #f0f0f0;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col0 {\n",
       "            background-color:  #727272;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col1 {\n",
       "            background-color:  #565656;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col2 {\n",
       "            background-color:  #3c3c3c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col3 {\n",
       "            background-color:  #e8e8e8;\n",
       "            color:  #000000;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col0 {\n",
       "            background-color:  #3b3b3b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col1 {\n",
       "            background-color:  #626262;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col2 {\n",
       "            background-color:  #515151;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col3 {\n",
       "            background-color:  #e6e6e6;\n",
       "            color:  #000000;\n",
       "        }</style><table id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >alt.atheism</th>        <th class=\"col_heading level0 col1\" >comp.graphics</th>        <th class=\"col_heading level0 col2\" >sci.space</th>        <th class=\"col_heading level0 col3\" >talk.religion.misc</th>    </tr></thead><tbody>\n",
       "                <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row0\" class=\"row_heading level0 row0\" >bobby</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col0\" class=\"data row0 col0\" >0.6674</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col1\" class=\"data row0 col1\" >-0.1608</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col2\" class=\"data row0 col2\" >-0.233</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row0_col3\" class=\"data row0 col3\" >-0.3152</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row1\" class=\"row_heading level0 row1\" >atheists</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col0\" class=\"data row1 col0\" >0.6578</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col1\" class=\"data row1 col1\" >-0.0889</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col2\" class=\"data row1 col2\" >-0.2178</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row1_col3\" class=\"data row1 col3\" >-0.4682</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row2\" class=\"row_heading level0 row2\" >atheism</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col0\" class=\"data row2 col0\" >0.6576</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col1\" class=\"data row2 col1\" >-0.2877</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col2\" class=\"data row2 col2\" >-0.2898</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row2_col3\" class=\"data row2 col3\" >-0.3274</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row3\" class=\"row_heading level0 row3\" >religion</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col0\" class=\"data row3 col0\" >0.6575</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col1\" class=\"data row3 col1\" >-0.4103</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col2\" class=\"data row3 col2\" >-0.5368</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row3_col3\" class=\"data row3 col3\" >-0.0278</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row4\" class=\"row_heading level0 row4\" >islam</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col0\" class=\"data row4 col0\" >0.5627</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col1\" class=\"data row4 col1\" >-0.0963</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col2\" class=\"data row4 col2\" >-0.2298</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row4_col3\" class=\"data row4 col3\" >-0.2322</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row5\" class=\"row_heading level0 row5\" >graphics</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col0\" class=\"data row5 col0\" >-0.535</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col1\" class=\"data row5 col1\" >1.3409</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col2\" class=\"data row5 col2\" >-0.8919</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row5_col3\" class=\"data row5 col3\" >-0.5064</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row6\" class=\"row_heading level0 row6\" >image</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col0\" class=\"data row6 col0\" >-0.3705</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col1\" class=\"data row6 col1\" >0.8997</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col2\" class=\"data row6 col2\" >-0.5303</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row6_col3\" class=\"data row6 col3\" >-0.3031</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row7\" class=\"row_heading level0 row7\" >file</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col0\" class=\"data row7 col0\" >-0.2274</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col1\" class=\"data row7 col1\" >0.8667</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col2\" class=\"data row7 col2\" >-0.5651</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row7_col3\" class=\"data row7 col3\" >-0.4027</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row8\" class=\"row_heading level0 row8\" >3d</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col0\" class=\"data row8 col0\" >-0.249</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col1\" class=\"data row8 col1\" >0.7653</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col2\" class=\"data row8 col2\" >-0.4505</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row8_col3\" class=\"data row8 col3\" >-0.2569</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row9\" class=\"row_heading level0 row9\" >computer</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col0\" class=\"data row9 col0\" >0.0109</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col1\" class=\"data row9 col1\" >0.7186</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col2\" class=\"data row9 col2\" >-0.4606</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row9_col3\" class=\"data row9 col3\" >-0.3169</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row10\" class=\"row_heading level0 row10\" >space</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col0\" class=\"data row10 col0\" >-0.8757</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col1\" class=\"data row10 col1\" >-0.9308</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col2\" class=\"data row10 col2\" >1.5896</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row10_col3\" class=\"data row10 col3\" >-0.7938</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row11\" class=\"row_heading level0 row11\" >orbit</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col0\" class=\"data row11 col0\" >-0.2902</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col1\" class=\"data row11 col1\" >-0.456</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col2\" class=\"data row11 col2\" >0.82</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row11_col3\" class=\"data row11 col3\" >-0.38</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row12\" class=\"row_heading level0 row12\" >nasa</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col0\" class=\"data row12 col0\" >-0.3727</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col1\" class=\"data row12 col1\" >-0.3431</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col2\" class=\"data row12 col2\" >0.7075</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row12_col3\" class=\"data row12 col3\" >-0.3361</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row13\" class=\"row_heading level0 row13\" >launch</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col0\" class=\"data row13 col0\" >-0.2972</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col1\" class=\"data row13 col1\" >-0.3295</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col2\" class=\"data row13 col2\" >0.6434</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row13_col3\" class=\"data row13 col3\" >-0.2302</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row14\" class=\"row_heading level0 row14\" >spacecraft</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col0\" class=\"data row14 col0\" >-0.2412</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col1\" class=\"data row14 col1\" >-0.258</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col2\" class=\"data row14 col2\" >0.5727</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row14_col3\" class=\"data row14 col3\" >-0.2171</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row15\" class=\"row_heading level0 row15\" >christian</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col0\" class=\"data row15 col0\" >-0.3749</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col1\" class=\"data row15 col1\" >-0.269</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col2\" class=\"data row15 col2\" >-0.2281</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row15_col3\" class=\"data row15 col3\" >0.749</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row16\" class=\"row_heading level0 row16\" >christians</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col0\" class=\"data row16 col0\" >-0.4813</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col1\" class=\"data row16 col1\" >-0.2397</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col2\" class=\"data row16 col2\" >-0.3048</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row16_col3\" class=\"data row16 col3\" >0.728</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row17\" class=\"row_heading level0 row17\" >blood</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col0\" class=\"data row17 col0\" >-0.3276</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col1\" class=\"data row17 col1\" >-0.0835</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col2\" class=\"data row17 col2\" >-0.1621</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row17_col3\" class=\"data row17 col3\" >0.6547</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row18\" class=\"row_heading level0 row18\" >fbi</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col0\" class=\"data row18 col0\" >-0.1865</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col1\" class=\"data row18 col1\" >-0.1668</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col2\" class=\"data row18 col2\" >-0.3054</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row18_col3\" class=\"data row18 col3\" >0.6044</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6level0_row19\" class=\"row_heading level0 row19\" >order</th>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col0\" class=\"data row19 col0\" >-0.5146</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col1\" class=\"data row19 col1\" >-0.0527</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col2\" class=\"data row19 col2\" >-0.1039</td>\n",
       "                        <td id=\"T_01e1b652_498e_11e9_b8ca_1002b51cfaa6row19_col3\" class=\"data row19 col3\" >0.5975</td>\n",
       "            </tr>\n",
       "    </tbody></table>"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x245c77fe5f8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part b - Logistic Regression - BiWord Vocabulary C = 0.25"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*Table of the top 5 features by weights per label*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<style  type=\"text/css\" >\n",
       "    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col1 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col2 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col3 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col4 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col1 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col2 {\n",
       "            background-color:  #444444;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col3 {\n",
       "            background-color:  #404040;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col4 {\n",
       "            background-color:  #656565;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col1 {\n",
       "            background-color:  #ededed;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col2 {\n",
       "            background-color:  #5f5f5f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col3 {\n",
       "            background-color:  #191919;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col4 {\n",
       "            background-color:  #666666;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col1 {\n",
       "            background-color:  #ebebeb;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col2 {\n",
       "            background-color:  #343434;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col3 {\n",
       "            background-color:  #1a1a1a;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col4 {\n",
       "            background-color:  #8e8e8e;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col1 {\n",
       "            background-color:  #ebebeb;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col2 {\n",
       "            background-color:  #3c3c3c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col3 {\n",
       "            background-color:  #595959;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col4 {\n",
       "            background-color:  #3f3f3f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col1 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col2 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col3 {\n",
       "            background-color:  #181818;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col4 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col1 {\n",
       "            background-color:  #242424;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col2 {\n",
       "            background-color:  #dadada;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col3 {\n",
       "            background-color:  #222222;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col4 {\n",
       "            background-color:  #212121;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col1 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col2 {\n",
       "            background-color:  #cecece;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col3 {\n",
       "            background-color:  #1b1b1b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col4 {\n",
       "            background-color:  #414141;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col1 {\n",
       "            background-color:  #424242;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col2 {\n",
       "            background-color:  #c9c9c9;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col3 {\n",
       "            background-color:  #1d1d1d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col4 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col1 {\n",
       "            background-color:  #4d4d4d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col2 {\n",
       "            background-color:  #c8c8c8;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col3 {\n",
       "            background-color:  #2f2f2f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col4 {\n",
       "            background-color:  #474747;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col1 {\n",
       "            background-color:  #4d4d4d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col2 {\n",
       "            background-color:  #151515;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col3 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col4 {\n",
       "            background-color:  #404040;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col1 {\n",
       "            background-color:  #393939;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col2 {\n",
       "            background-color:  #161616;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col3 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col4 {\n",
       "            background-color:  #4b4b4b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col1 {\n",
       "            background-color:  #525252;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col2 {\n",
       "            background-color:  #2f2f2f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col3 {\n",
       "            background-color:  #d4d4d4;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col4 {\n",
       "            background-color:  #4f4f4f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col1 {\n",
       "            background-color:  #5c5c5c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col2 {\n",
       "            background-color:  #2e2e2e;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col3 {\n",
       "            background-color:  #d1d1d1;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col4 {\n",
       "            background-color:  #4d4d4d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col1 {\n",
       "            background-color:  #5f5f5f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col2 {\n",
       "            background-color:  #313131;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col3 {\n",
       "            background-color:  #c6c6c6;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col4 {\n",
       "            background-color:  #3c3c3c;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col1 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col2 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col3 {\n",
       "            background-color:  #000000;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col4 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col1 {\n",
       "            background-color:  #6d6d6d;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col2 {\n",
       "            background-color:  #424242;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col3 {\n",
       "            background-color:  #3b3b3b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col4 {\n",
       "            background-color:  #ffffff;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col1 {\n",
       "            background-color:  #646464;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col2 {\n",
       "            background-color:  #424242;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col3 {\n",
       "            background-color:  #565656;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col4 {\n",
       "            background-color:  #eaeaea;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col1 {\n",
       "            background-color:  #595959;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col2 {\n",
       "            background-color:  #6b6b6b;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col3 {\n",
       "            background-color:  #323232;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col4 {\n",
       "            background-color:  #e5e5e5;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col1 {\n",
       "            background-color:  #979797;\n",
       "            color:  #000000;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col2 {\n",
       "            background-color:  #3b3b3b;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col3 {\n",
       "            background-color:  #3d3d3d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col4 {\n",
       "            background-color:  #e1e1e1;\n",
       "            color:  #000000;\n",
       "        }</style><table id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >index</th>        <th class=\"col_heading level0 col1\" >alt.atheism</th>        <th class=\"col_heading level0 col2\" >comp.graphics</th>        <th class=\"col_heading level0 col3\" >sci.space</th>        <th class=\"col_heading level0 col4\" >talk.religion.misc</th>    </tr></thead><tbody>\n",
       "                <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row0\" class=\"row_heading level0 row0\" >0</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col0\" class=\"data row0 col0\" >cheers kent</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col1\" class=\"data row0 col1\" >0.4526</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col2\" class=\"data row0 col2\" >-0.5413</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col3\" class=\"data row0 col3\" >-0.521</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row0_col4\" class=\"data row0 col4\" >0.4533</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row1\" class=\"row_heading level0 row1\" >1</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col0\" class=\"data row1 col0\" >claim that</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col1\" class=\"data row1 col1\" >0.4499</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col2\" class=\"data row1 col2\" >-0.1511</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col3\" class=\"data row1 col3\" >-0.209</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row1_col4\" class=\"data row1 col4\" >-0.0958</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row2\" class=\"row_heading level0 row2\" >2</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col0\" class=\"data row2 col0\" >in this</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col1\" class=\"data row2 col1\" >0.3828</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col2\" class=\"data row2 col2\" >-0.0024</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col3\" class=\"data row2 col3\" >-0.3971</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row2_col4\" class=\"data row2 col4\" >-0.093</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row3\" class=\"row_heading level0 row3\" >3</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col0\" class=\"data row3 col0\" >you are</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col1\" class=\"data row3 col1\" >0.3767</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col2\" class=\"data row3 col2\" >-0.2412</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col3\" class=\"data row3 col3\" >-0.3951</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row3_col4\" class=\"data row3 col4\" >0.0497</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row4\" class=\"row_heading level0 row4\" >4</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col0\" class=\"data row4 col0\" >are you</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col1\" class=\"data row4 col1\" >0.3735</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col2\" class=\"data row4 col2\" >-0.2005</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col3\" class=\"data row4 col3\" >-0.0878</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row4_col4\" class=\"data row4 col4\" >-0.2322</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row5\" class=\"row_heading level0 row5\" >5</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col0\" class=\"data row5 col0\" >looking for</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col1\" class=\"data row5 col1\" >-0.5173</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col2\" class=\"data row5 col2\" >0.9089</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col3\" class=\"data row5 col3\" >-0.402</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row5_col4\" class=\"data row5 col4\" >-0.4586</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row6\" class=\"row_heading level0 row6\" >6</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col0\" class=\"data row6 col0\" >in advance</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col1\" class=\"data row6 col1\" >-0.3799</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col2\" class=\"data row6 col2\" >0.696</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col3\" class=\"data row6 col3\" >-0.3552</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row6_col4\" class=\"data row6 col4\" >-0.339</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row7\" class=\"row_heading level0 row7\" >7</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col0\" class=\"data row7 col0\" >out there</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col1\" class=\"data row7 col1\" >-0.2295</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col2\" class=\"data row7 col2\" >0.6262</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col3\" class=\"data row7 col3\" >-0.3864</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row7_col4\" class=\"data row7 col4\" >-0.2257</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row8\" class=\"row_heading level0 row8\" >8</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col0\" class=\"data row8 col0\" >is there</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col1\" class=\"data row8 col1\" >-0.2637</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col2\" class=\"data row8 col2\" >0.6009</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col3\" class=\"data row8 col3\" >-0.3785</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row8_col4\" class=\"data row8 col4\" >-0.1898</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row9\" class=\"row_heading level0 row9\" >9</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col0\" class=\"data row9 col0\" >comp graphics</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col1\" class=\"data row9 col1\" >-0.2226</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col2\" class=\"data row9 col2\" >0.595</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col3\" class=\"data row9 col3\" >-0.2898</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row9_col4\" class=\"data row9 col4\" >-0.2028</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row10\" class=\"row_heading level0 row10\" >10</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col0\" class=\"data row10 col0\" >the space</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col1\" class=\"data row10 col1\" >-0.2253</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col2\" class=\"data row10 col2\" >-0.4206</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col3\" class=\"data row10 col3\" >0.7166</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row10_col4\" class=\"data row10 col4\" >-0.2274</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row11\" class=\"row_heading level0 row11\" >11</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col0\" class=\"data row11 col0\" >the moon</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col1\" class=\"data row11 col1\" >-0.2981</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col2\" class=\"data row11 col2\" >-0.4113</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col3\" class=\"data row11 col3\" >0.7139</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row11_col4\" class=\"data row11 col4\" >-0.189</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row12\" class=\"row_heading level0 row12\" >12</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col0\" class=\"data row12 col0\" >sci space</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col1\" class=\"data row12 col1\" >-0.2057</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col2\" class=\"data row12 col2\" >-0.2727</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col3\" class=\"data row12 col3\" >0.5067</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row12_col4\" class=\"data row12 col4\" >-0.175</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row13\" class=\"row_heading level0 row13\" >13</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col0\" class=\"data row13 col0\" >and such</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col1\" class=\"data row13 col1\" >-0.1682</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col2\" class=\"data row13 col2\" >-0.2775</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col3\" class=\"data row13 col3\" >0.494</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row13_col4\" class=\"data row13 col4\" >-0.1813</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row14\" class=\"row_heading level0 row14\" >14</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col0\" class=\"data row14 col0\" >it was</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col1\" class=\"data row14 col1\" >-0.1548</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col2\" class=\"data row14 col2\" >-0.2586</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col3\" class=\"data row14 col3\" >0.4394</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row14_col4\" class=\"data row14 col4\" >-0.2447</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row15\" class=\"row_heading level0 row15\" >15</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col0\" class=\"data row15 col0\" >cheers kent</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col1\" class=\"data row15 col1\" >0.4526</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col2\" class=\"data row15 col2\" >-0.5413</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col3\" class=\"data row15 col3\" >-0.521</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row15_col4\" class=\"data row15 col4\" >0.4533</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row16\" class=\"row_heading level0 row16\" >16</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col0\" class=\"data row16 col0\" >the fbi</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col1\" class=\"data row16 col1\" >-0.1025</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col2\" class=\"data row16 col2\" >-0.1657</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col3\" class=\"data row16 col3\" >-0.2352</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row16_col4\" class=\"data row16 col4\" >0.4504</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row17\" class=\"row_heading level0 row17\" >17</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col0\" class=\"data row17 col0\" >but he</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col1\" class=\"data row17 col1\" >-0.136</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col2\" class=\"data row17 col2\" >-0.1654</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col3\" class=\"data row17 col3\" >-0.1021</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row17_col4\" class=\"data row17 col4\" >0.3764</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row18\" class=\"row_heading level0 row18\" >18</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col0\" class=\"data row18 col0\" >with you</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col1\" class=\"data row18 col1\" >-0.1765</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col2\" class=\"data row18 col2\" >0.07</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col3\" class=\"data row18 col3\" >-0.2768</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row18_col4\" class=\"data row18 col4\" >0.36</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6level0_row19\" class=\"row_heading level0 row19\" >19</th>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col0\" class=\"data row19 col0\" >the word</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col1\" class=\"data row19 col1\" >0.0566</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col2\" class=\"data row19 col2\" >-0.2027</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col3\" class=\"data row19 col3\" >-0.2222</td>\n",
       "                        <td id=\"T_03cf8a52_498e_11e9_976d_1002b51cfaa6row19_col4\" class=\"data row19 col4\" >0.3449</td>\n",
       "            </tr>\n",
       "    </tbody></table>"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x245ca6eb7b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P4():\n",
    "    \"\"\" \n",
    "    4th Problem. Using the default features vector, train a Logistic Regression, using the C from P3. \n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment.\n",
    "    Returns:\n",
    "        The function will return a report with two outputs: \n",
    "            a.- Table with 20 rows and 4 columns of the weights of each feature using standard Vocabularory\n",
    "            b.- Similar table using bigram features instead\n",
    "    \"\"\"   \n",
    "    \n",
    "    # Part a - Standard Vocabulary. Optimal C=0.25\n",
    "    classes = ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n",
    "    log_reg_clf = Pipeline([\n",
    "        ('vect', CountVectorizer()),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=0.25, n_jobs=-1)),\n",
    "    ])\n",
    "    log_reg_clf.fit(train_data, train_labels)\n",
    "    top_5_indices = np.fliplr(log_reg_clf.named_steps['clf'].coef_.argsort()[:,-5:]).flatten()\n",
    "    features = np.array(log_reg_clf.named_steps['vect'].get_feature_names())[top_5_indices]\n",
    "    weights = log_reg_clf.named_steps['clf'].coef_[:,top_5_indices].T\n",
    "    table_feat_weight = pd.DataFrame(weights, index=features, columns = classes)\n",
    "    printmd('### Part a - Logistic Regression - Word Vocabulary C = 0.25')\n",
    "    printmd('*Table of the top 5 features by weights per label*')\n",
    "    display(table_feat_weight.round(4).style.background_gradient(cmap='gist_gray'))\n",
    "    \n",
    "    # Part b - Bigram Vocabulary. Optimal C=0.25\n",
    "    classes = ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n",
    "    log_reg_clf_bi = Pipeline([\n",
    "        ('vect', CountVectorizer(analyzer='word', ngram_range=(2,2))),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=0.25, n_jobs=-1)),\n",
    "    ])\n",
    "    log_reg_clf_bi.fit(train_data, train_labels)\n",
    "    top_5_indices_bi = np.fliplr(log_reg_clf_bi.named_steps['clf'].coef_.argsort()[:,-5:]).flatten()\n",
    "    features_bi = np.array(log_reg_clf_bi.named_steps['vect'].get_feature_names())[top_5_indices_bi]\n",
    "    weights_bi = log_reg_clf_bi.named_steps['clf'].coef_[:,top_5_indices_bi].T\n",
    "    table_feat_weight_bi = pd.DataFrame(weights_bi, index=features_bi, columns = classes)\n",
    "    printmd('### Part b - Logistic Regression - BiWord Vocabulary C = 0.25')\n",
    "    printmd('*Table of the top 5 features by weights per label*')\n",
    "    display_df = table_feat_weight_bi.reset_index()\n",
    "    display(display_df.round(4).style.background_gradient(cmap='gist_gray'))\n",
    "\n",
    "P4()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Answer:\n",
    "\n",
    "Is quite surprising that when we increase the vocubaluray to bigram words, we start 'losing' some of the interpretability of the model. For a 1-word model, for each label, the most important features seem to be related to the space: religion, islam for atheism; graphics, image for comp.graphics; etc. However, when we increase to 2-Word models, the most important features start to be common combinations of words in the English language: 'with you', 'you are', 'are you', etc. Although there are combinations that seems quite on the spot like 'the moon' or 'sci space', the model start showing one of the main issues of text classification: how to model the English Language and all the possible combinations that happens naturally just because, the so called 'Stop Words' start to appear more often because of these combinations."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (5) \n",
    "Try to improve the logistic regression classifier by passing a custom preprocessor to CountVectorizer. The preprocessing function runs on the raw text, before it is split into words by the tokenizer. Your preprocessor should try to normalize the input in various ways to improve generalization. For example, try lowercasing everything, replacing sequences of numbers with a single token, removing various other non-letter characters, and shortening long words. If you're not already familiar with regular expressions for manipulating strings, see https://docs.python.org/2/library/re.html, and re.sub() in particular. With your new preprocessor, how much did you reduce the size of the dictionary?\n",
    "\n",
    "For reference, I was able to improve dev F1 by 2 points."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "## Training with Train Data, Evaluating Performance in Dev Data"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part a - Empty Preprocessor"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "The size of the vocabulary is 33291 and the F1 score is 0.710"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part b - Custom Preprocessor"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "The size of the vocabulary is 21750 and the F1 score is 0.714"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*Difference in words is 11541 and the increase(decrease) in F1 score is 0.0044*"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def empty_preprocessor(s):\n",
    "    return s\n",
    "\n",
    "def better_preprocessor(s):\n",
    "    \"\"\"\n",
    "    Function for preprocessing the string data for better classification\n",
    "    Args:\n",
    "        String\n",
    "    Returns:\n",
    "        Processed String\n",
    "    \"\"\"\n",
    "    # Let's start by lowering everything and doing Regex (non-letter, extra spaces, plural, numbers)\n",
    "    s = s.lower()\n",
    "    s_new = re.sub(r\"[,@\\'?\\.$%_]\", \"\", s, flags=re.I) \n",
    "    s_new = re.sub(r\"\\s+\",\" \", s_new, flags = re.I)  \n",
    "    s_new = re.sub(r'ies\\b|es\\b|s\\b|y\\b', '', s_new, flags=re.I)\n",
    "    s_new = re.sub(r'\\w*\\d\\w*|_', '', s_new).strip()\n",
    "    \n",
    "    # Now, let's use NLP for removing Stop Words and tokenize\n",
    "    stop_words = set(stopwords.words('english')) \n",
    "    word_tokens = word_tokenize(s_new) \n",
    "    filtered_s = [w for w in word_tokens if not w in stop_words] \n",
    "    filtered_s = [] \n",
    "    for w in word_tokens: \n",
    "        if w not in stop_words: \n",
    "            filtered_s.append(w) \n",
    "    return str(filtered_s)\n",
    "\n",
    "def P5():\n",
    "    \"\"\" \n",
    "    5th Problem. Comparing empty preprocessor with custom preprocessor to improve training and reduce the size of vocabulary\n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment. C used equal to 0.25 as before\n",
    "    Returns:\n",
    "        The function will return a report of the difference of sizes of the vocabularies and the F1 scores\n",
    "    \"\"\"   \n",
    "    # Part a - Empty Preprocessor\n",
    "    log_reg_clf_emp = Pipeline([\n",
    "        ('vect', CountVectorizer(preprocessor=empty_preprocessor)),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=0.25, n_jobs=-1)),\n",
    "    ])\n",
    "    log_reg_clf_emp.fit(train_data, train_labels)\n",
    "    printmd('## Training with Train Data, Evaluating Performance in Dev Data')\n",
    "    printmd('### Part a - Empty Preprocessor')\n",
    "    voc_emp = log_reg_clf_emp.named_steps['clf'].coef_.shape[1]\n",
    "    y_dev_pre_emp = log_reg_clf_emp.predict(dev_data)\n",
    "    f1_emp = metrics.f1_score(dev_labels, y_dev_pre_emp,average='micro')\n",
    "    printmd('The size of the vocabulary is %d and the F1 score is %.03f'%(voc_emp, f1_emp))\n",
    "    \n",
    "    # Part b - Custom Preprocessor\n",
    "    log_reg_clf_cst = Pipeline([\n",
    "        ('vect', CountVectorizer(preprocessor=better_preprocessor)),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=0.25, n_jobs=-1)),\n",
    "    ])\n",
    "    log_reg_clf_cst.fit(train_data, train_labels)\n",
    "    printmd('### Part b - Custom Preprocessor')\n",
    "    voc_cst = log_reg_clf_cst.named_steps['clf'].coef_.shape[1]\n",
    "    y_dev_pre_cst = log_reg_clf_cst.predict(dev_data)\n",
    "    f1_cst = metrics.f1_score(dev_labels, y_dev_pre_cst,average='micro')\n",
    "    printmd('The size of the vocabulary is %d and the F1 score is %.03f'%(voc_cst, f1_cst))\n",
    "    printmd('*The preprocessor reduced the vocabulary by %d and increased(decrease) the F1 score by %.04f*'%(voc_emp-\\\n",
    "                                                                                                voc_cst,f1_cst-f1_emp))\n",
    "    \n",
    "P5()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (6) \n",
    "The idea of regularization is to avoid learning very large weights (which are likely to fit the training data, but not generalize well) by adding a penalty to the total size of the learned weights. That is, logistic regression seeks the set of weights that minimizes errors in the training data AND has a small size. The default regularization, L2, computes this size as the sum of the squared weights (see P3, above). L1 regularization computes this size as the sum of the absolute values of the weights. The result is that whereas L2 regularization makes all the weights relatively small, L1 regularization drives lots of the weights to 0, effectively removing unimportant features.\n",
    "\n",
    "Train a logistic regression model using a \"l1\" penalty. Output the number of learned weights that are not equal to zero. How does this compare to the number of non-zero weights you get with \"l2\"? Now, reduce the size of the vocabulary by keeping only those features that have at least one non-zero weight and retrain a model using \"l2\".\n",
    "\n",
    "Make a plot showing accuracy of the re-trained model vs. the vocabulary size you get when pruning unused features by adjusting the C parameter.\n",
    "\n",
    "Note: The gradient descent code that trains the logistic regression model sometimes has trouble converging with extreme settings of the C parameter. Relax the convergence criteria by setting tol=.01 (the default is .0001)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA8kAAAEsCAYAAADw2vGBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245c78501d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P6():\n",
    "    \"\"\" \n",
    "    6th Problem. Use L1 Regularization for feature selection\n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment. C will be the same for both models\n",
    "    Returns:\n",
    "        Plot of the size of the vocabulary vs. the accuracy of the model. \n",
    "    \"\"\"   \n",
    "    # Parameters of the iteration and Gradient Descent\n",
    "    np.random.seed(0)\n",
    "    C = [0.01, 0.1, 0.2, 0.5, 0.8, 1, 1.5, 2.0, 5.0, 10.0, 15.0, 20.0]\n",
    "    tol = 0.01\n",
    "    size_voc = []\n",
    "    acc = []\n",
    "    for c in C:\n",
    "        # L1 Regularization for feature selection\n",
    "        log_reg_clf_l1 = Pipeline([\n",
    "            ('vect', CountVectorizer()),\n",
    "            ('clf', LogisticRegression(penalty='l1', C=c, n_jobs=-1, tol=0.01)),\n",
    "        ])\n",
    "        log_reg_clf_l1.fit(train_data, train_labels)\n",
    "        log_reg_clf_l1_coef = np.sum(log_reg_clf_l1.named_steps['clf'].coef_,axis=0).nonzero()\n",
    "        log_reg_clf_l1_vocab = log_reg_clf_l1.named_steps['vect'].get_feature_names()\n",
    "        l1_features = np.array(log_reg_clf_l1_vocab)[log_reg_clf_l1_coef]\n",
    "        \n",
    "        # L2 Regularization for Training\n",
    "        log_reg_clf_l2 = Pipeline([\n",
    "            ('vect', CountVectorizer(vocabulary=l1_features)),\n",
    "            ('clf', LogisticRegression(penalty='l2', C=c, n_jobs=-1, tol=0.01)),\n",
    "        ])\n",
    "        log_reg_clf_l2.fit(train_data, train_labels)\n",
    "        size_voc.append(log_reg_clf_l2.named_steps['clf'].coef_.shape[1])\n",
    "        y_dev_pre_cst = log_reg_clf_l2.predict(dev_data)\n",
    "        acc.append(metrics.accuracy_score(dev_labels, y_dev_pre_cst))\n",
    "    \n",
    "    # Plot the results\n",
    "    fig, ax = plt.subplots(figsize=(16,4))\n",
    "    cmap = matplotlib.cm.get_cmap('viridis')\n",
    "    normalize = matplotlib.colors.Normalize(vmin=min(C), vmax=max(C))\n",
    "    colors = [cmap(normalize(value)) for value in C]\n",
    "    ax.scatter(size_voc, acc, marker='o', color=colors)\n",
    "    ax.set_title('Accuracy vs. Vocabulary size: Feature Selection using L1 Regularization')\n",
    "    ax.set_xlabel('Vocabulary Size')\n",
    "    ax.set_ylabel('Accuracy in Dev Set')\n",
    "    cax, _ = matplotlib.colorbar.make_axes(ax)\n",
    "    cbar = matplotlib.colorbar.ColorbarBase(cax, cmap=cmap, norm=normalize)\n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "\n",
    "P6()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Answer\n",
    "\n",
    "It's quite clear how L1 regularization shrinks the vocabulary size in quite a spectacular way. This is a strategy also use in LASSO regression, where we do feature selection first by using L1 norm. The reasoning behind it is that L1 ignores any kind of outliers of the data, and tends to the median, instead of the mean. Thus, the weights are 'solved' to zero quickly by the linear programming. On the other hand, L2 norm tends to the average, thus outliers have a huge impact on it. The way L2 handles it is to have almost all the weigths with at least some value."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (7)\n",
    "Use the TfidfVectorizer -- how is this different from the CountVectorizer? Train a logistic regression model with C=100.\n",
    "\n",
    "Make predictions on the dev data and show the top 3 documents where the ratio R is largest, where R is:\n",
    "\n",
    "maximum predicted probability / predicted probability of the correct label\n",
    "\n",
    "What kinds of mistakes is the model making? Suggest a way to address one particular issue that you see."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "**Documents with largest R ratio**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 273.82**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 1 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*talk.religion.misc - comp.graphics*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 1**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Can anyone provide me a ftp site where I can obtain a online version\n",
      "of the Book of Mormon. Please email the internet address if possible.\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 639.31**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 2 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*alt.atheism - talk.religion.misc*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 2**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "The 24 children were, of course, killed by a lone gunman in a second story\n",
      "window, who fired eight bullets in the space of two seconds...\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 1552.17**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 3 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*talk.religion.misc - comp.graphics*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 3**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I am pleased to announce that a *revised version* of _The Easy-to-Read Book\n",
      "of Mormon_ (former title: _Mormon's Book_) by Lynn Matthews Anderson is now\n",
      "available through anonymous ftp (see information below). In addition to the\n",
      "change in title, the revised ETR BOM has been shortened by several pages\n",
      "(eliminating many extraneous \"that's\" and \"of's\"), and many (minor) errors\n",
      "have been corrected. This release includes a simplified Joseph Smith Story,\n",
      "testimonies of the three and eight witnesses, and a \"Words-to-Know\"\n",
      "glossary.\n",
      "\n",
      "As with the previous announcement, readers are reminded that this is a\n",
      "not-for-profit endeavor. This is a copyrighted work, but people are welcome\n",
      "to make *verbatim* copies for personal use. People can recuperate the\n",
      "actual costs of printing (paper, copy center charges), but may not charge\n",
      "anything for their time in making copies, or in any way realize a profit\n",
      "from the use of this book. See the permissions notice in the book itself\n",
      "for the precise terms.\n",
      "\n",
      "Negotiations are currently underway with a Mormon publisher vis-a-vis the\n",
      "printing and distribution of bound books. (Sorry, I'm out of the wire-bound\n",
      "\"first editions.\") I will make another announcement about the availability\n",
      "of printed copies once everything has been worked out.\n",
      "\n",
      "FTP information: connect via anonymous ftp to carnot.itc.cmu.edu, then \"cd\n",
      "pub\" (you won't see anything at all until you do).\n",
      "\n",
      "\"The Easy-to-Read Book of Mormon\" is currently available in postscript and\n",
      "RTF (rich text format). (ASCII, LaTeX, and other versions can be made\n",
      "available; contact dba@andrew.cmu.edu for details.) You should be able to\n",
      "print the postscript file on any postscript printer (such as an Apple\n",
      "Laserwriter); let dba know if you have any difficulties. (The postscript in\n",
      "the last release had problems on some printers; this time it should work\n",
      "better.) RTF is a standard document interchange format that can be read in\n",
      "by a number of word processors, including Microsoft Word for both the\n",
      "Macintosh and Windows. If you don't have a postscript printer, you may be\n",
      "able to use the RTF file to print out a copy of the book.\n",
      "\n",
      "-r--r--r--  1 dba                   1984742 Apr 27 13:12 etrbom.ps\n",
      "-r--r--r--  1 dba                   1209071 Apr 27 13:13 etrbom.rtf\n",
      "\n",
      "For more information about how this project came about, please refer to my\n",
      "article in the current issue of _Sunstone_, entitled \"Delighting in\n",
      "Plainness: Issues Surrounding a Simple Modern English Book of Mormon.\"\n",
      "\n",
      "Send all inquiries and comments to:\n",
      "\n",
      "    Lynn Matthews Anderson\n",
      "    5806 Hampton Street\n",
      "    Pittsburgh, PA 15206\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABD4AAAEsCAYAAAArcXPiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x245afd1f1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P7():\n",
    "    \"\"\" \n",
    "    7th Problem. Introduce TdifVectorizer to the Pipeline\n",
    "    Args:\n",
    "        No arguments. All data is loaded in the global environment. C=100\n",
    "    Returns:\n",
    "        Report the 3 documents with the highest R ratio  \n",
    "    \"\"\" \n",
    "    # Train the classifier\n",
    "    classes = ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n",
    "    log_reg_clf_final = Pipeline([\n",
    "        ('vect', CountVectorizer(preprocessor=better_preprocessor)),\n",
    "        ('tdif',TfidfTransformer()),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=100, n_jobs=-1)),\n",
    "    ])\n",
    "    log_reg_clf_final.fit(train_data, train_labels)\n",
    "    y_dev_pre_final = log_reg_clf_final.predict(dev_data)\n",
    "    prob = log_reg_clf_final.predict_proba(dev_data)\n",
    "    correct_label_prob = np.zeros(676)\n",
    "    for i in range(dev_labels.shape[0]):\n",
    "        correct_label_prob[i] = prob[i,dev_labels[i]]\n",
    "    R = np.max(prob, axis=1)/correct_label_prob\n",
    "    max_R_index = R.argsort()[-3:]\n",
    "    printmd(\"**Documents with largest R ratio**\")\n",
    "    j = 1\n",
    "    for i in max_R_index:\n",
    "        printmd('**-**'*100+'\\n')\n",
    "        printmd('**Value or R is %.2f**'%R[i])\n",
    "        printmd('**Label of development example %d and label predicted**'%(j))\n",
    "        target_name = classes [y_dev_pre_final[i]]\n",
    "        printmd('*%s - %s*'%(newsgroups_test.target_names[dev_labels[i]],target_name)+'\\n')\n",
    "        printmd('**Text of the message in training example %d**'%(j))\n",
    "        print(\"\\n\".join(dev_data[i].split(\"\\n\")))\n",
    "        printmd('**-**'*100+'\\n')\n",
    "        j += 1\n",
    "        \n",
    "    # Plot the length of each message vs. R\n",
    "    len_s = [len(s) for s in dev_data]\n",
    "    fig, ax = plt.subplots(figsize=(16,4))\n",
    "    ax.scatter(R, len_s, marker='o')\n",
    "    ax.set_title('Length of message vs. R in the Development Set')\n",
    "    ax.set_xlabel('R')\n",
    "    ax.set_ylabel('Length of Message')\n",
    "    plt.show()\n",
    "    \n",
    "P7()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Answer:\n",
    "\n",
    "For these three messages we can infer why the classifier had problems: for the first and last, the message encoded was describing how to get a certain book online, with a quite specific description on how to download it, formats, etc. These words match with the vocabulary of the label predicted more than the actual label, thus the problem. For the second message, it appears it's quite short and there is something that's confusing the classifier. \n",
    "\n",
    "Occurrence count is a good start but there is an issue: longer documents will have higher average count values than shorter documents, even though they might talk about the same topics. To avoid these potential discrepancies it suffices to divide the number of occurrences of each word in a document by the total number of words in the document: these new features are called tf for Term Frequencies"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## (8) EXTRA CREDIT\n",
    "\n",
    "Try implementing one of your ideas based on your error analysis. Use logistic regression as your underlying model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "### Part a - Improvement in Accuracy"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "New accuracy of this model is 0.740"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "### Part b - Examples of misslabeling"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Documents with largest R ratio in new model**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 520.37**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 1 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*alt.atheism - comp.graphics*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 1**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "The CLIPPER initiative is an announcement by Clinton that all the \n",
      "\"secure\" voice phones will use the same crypto chip, as a de-facto\n",
      "government standard.  Problem is, the government is admitting that\n",
      "they hold the keys to break the code easily, and the Justice department\n",
      "will be using the keys to listen in on \"illegal activities.\"  Many\n",
      "people are really scared about such an initiative because it is\n",
      "a major step towards outlawing real crypto protection on things\n",
      "like email if you read the press release.  The project was developed\n",
      "by NSA and given to NIST.  It uses two keys S1 and S2 that the\n",
      "government claims are needed to break the code.  They claim that\n",
      "these keys will be handed to two different companies, and when they\n",
      "get a warrant to do a wiretap (the chip is nicknamed the wiretap chip),\n",
      "they have to get the keys from both companies.  People have poked holes\n",
      "through and through the press release official version and shown how\n",
      "it is nowhere near as nice as it sounds, and I have given the simplified\n",
      "version.  People over on sci.crypt are really scared about this\n",
      "proposal it seems.\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 783.16**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 2 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*talk.religion.misc - comp.graphics*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 2**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I am pleased to announce that a *revised version* of _The Easy-to-Read Book\n",
      "of Mormon_ (former title: _Mormon's Book_) by Lynn Matthews Anderson is now\n",
      "available through anonymous ftp (see information below). In addition to the\n",
      "change in title, the revised ETR BOM has been shortened by several pages\n",
      "(eliminating many extraneous \"that's\" and \"of's\"), and many (minor) errors\n",
      "have been corrected. This release includes a simplified Joseph Smith Story,\n",
      "testimonies of the three and eight witnesses, and a \"Words-to-Know\"\n",
      "glossary.\n",
      "\n",
      "As with the previous announcement, readers are reminded that this is a\n",
      "not-for-profit endeavor. This is a copyrighted work, but people are welcome\n",
      "to make *verbatim* copies for personal use. People can recuperate the\n",
      "actual costs of printing (paper, copy center charges), but may not charge\n",
      "anything for their time in making copies, or in any way realize a profit\n",
      "from the use of this book. See the permissions notice in the book itself\n",
      "for the precise terms.\n",
      "\n",
      "Negotiations are currently underway with a Mormon publisher vis-a-vis the\n",
      "printing and distribution of bound books. (Sorry, I'm out of the wire-bound\n",
      "\"first editions.\") I will make another announcement about the availability\n",
      "of printed copies once everything has been worked out.\n",
      "\n",
      "FTP information: connect via anonymous ftp to carnot.itc.cmu.edu, then \"cd\n",
      "pub\" (you won't see anything at all until you do).\n",
      "\n",
      "\"The Easy-to-Read Book of Mormon\" is currently available in postscript and\n",
      "RTF (rich text format). (ASCII, LaTeX, and other versions can be made\n",
      "available; contact dba@andrew.cmu.edu for details.) You should be able to\n",
      "print the postscript file on any postscript printer (such as an Apple\n",
      "Laserwriter); let dba know if you have any difficulties. (The postscript in\n",
      "the last release had problems on some printers; this time it should work\n",
      "better.) RTF is a standard document interchange format that can be read in\n",
      "by a number of word processors, including Microsoft Word for both the\n",
      "Macintosh and Windows. If you don't have a postscript printer, you may be\n",
      "able to use the RTF file to print out a copy of the book.\n",
      "\n",
      "-r--r--r--  1 dba                   1984742 Apr 27 13:12 etrbom.ps\n",
      "-r--r--r--  1 dba                   1209071 Apr 27 13:13 etrbom.rtf\n",
      "\n",
      "For more information about how this project came about, please refer to my\n",
      "article in the current issue of _Sunstone_, entitled \"Delighting in\n",
      "Plainness: Issues Surrounding a Simple Modern English Book of Mormon.\"\n",
      "\n",
      "Send all inquiries and comments to:\n",
      "\n",
      "    Lynn Matthews Anderson\n",
      "    5806 Hampton Street\n",
      "    Pittsburgh, PA 15206\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Value or R is 1135.22**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Label of development example 3 and label predicted**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "*alt.atheism - talk.religion.misc*\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "**Text of the message in training example 3**"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "The 24 children were, of course, killed by a lone gunman in a second story\n",
      "window, who fired eight bullets in the space of two seconds...\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "**-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-****-**\n"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def P8():\n",
    "    \"\"\" \n",
    "    8th Problem. Improve the accuracy of logistic regression\n",
    "    Args:\n",
    "        All data is loaded in the global environment. From P6, I'll use C=20\n",
    "    Returns:\n",
    "        Prints in console accuracy of improvement\n",
    "    \"\"\"  \n",
    "    # Selection of features using L1 Regularization and TDIf\n",
    "    classes = ['alt.atheism', 'comp.graphics', 'sci.space', 'talk.religion.misc']\n",
    "    log_reg_clf_l1 = Pipeline([\n",
    "        ('vect', CountVectorizer(preprocessor=better_preprocessor, min_df=2, max_df=0.12, stop_words='english')),\n",
    "        ('tdif', TfidfTransformer(norm='l1')),\n",
    "        ('clf', LogisticRegression(penalty='l1', C=20, n_jobs=-1, tol=0.001)),\n",
    "    ])\n",
    "    log_reg_clf_l1.fit(train_data, train_labels)\n",
    "    log_reg_clf_l1_coef = np.sum(log_reg_clf_l1.named_steps['clf'].coef_,axis=0).nonzero()\n",
    "    log_reg_clf_l1_vocab = log_reg_clf_l1.named_steps['vect'].get_feature_names()\n",
    "    l1_features = np.array(log_reg_clf_l1_vocab)[log_reg_clf_l1_coef]\n",
    "    \n",
    "    # Train using L2 and vocabulary\n",
    "    log_reg_clf_l2 = Pipeline([\n",
    "        ('vect', CountVectorizer(preprocessor=better_preprocessor, vocabulary=l1_features,\\\n",
    "                                 min_df=2, max_df=0.2, stop_words='english')),\n",
    "        ('tdif', TfidfTransformer()),\n",
    "        ('clf', LogisticRegression(penalty='l2', C=20, n_jobs=-1, tol=0.001)),\n",
    "    ])\n",
    "    log_reg_clf_l2.fit(train_data, train_labels)\n",
    "    y_dev_pre = log_reg_clf_l2.predict(dev_data)\n",
    "    printmd('### Part a - Improvement in Accuracy')\n",
    "    printmd('New accuracy of this model is %.03f'%metrics.accuracy_score(dev_labels, y_dev_pre))\n",
    "    prob = log_reg_clf_l2.predict_proba(dev_data)\n",
    "    correct_label_prob = np.zeros(676)\n",
    "    for i in range(dev_labels.shape[0]):\n",
    "        correct_label_prob[i] = prob[i,dev_labels[i]]\n",
    "    R = np.max(prob, axis=1)/correct_label_prob\n",
    "    max_R_index = R.argsort()[-3:]\n",
    "    printmd('### Part b - Examples of misslabeling')\n",
    "    printmd(\"**Documents with largest R ratio in new model**\")\n",
    "    j = 1\n",
    "    for i in max_R_index:\n",
    "        printmd('**-**'*100+'\\n')\n",
    "        printmd('**Value or R is %.2f**'%R[i])\n",
    "        printmd('**Label of development example %d and label predicted**'%(j))\n",
    "        target_name = classes [y_dev_pre[i]]\n",
    "        printmd('*%s - %s*'%(newsgroups_test.target_names[dev_labels[i]],target_name)+'\\n')\n",
    "        printmd('**Text of the message in training example %d**'%(j))\n",
    "        print(\"\\n\".join(dev_data[i].split(\"\\n\")))\n",
    "        printmd('**-**'*100+'\\n')\n",
    "        j += 1\n",
    "        \n",
    "P8()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Answer:\n",
    "\n",
    "With an ensemble tactic we are able to improve our accuracy by 4%, all next work should be focus on preprocessing the data and creating dictionaries for each label, to improve the vocabularies (done here using L1 regularization)\n",
    "\n",
    "Also, although we have two similar occurrances of mislabel as before, all three mislabels can be attributed to the similarity of the languages and the vocabularies of the training. With more data, we can improve our performance by improving the corpus of each label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}