ReadingData.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:df018acebea0c8f4c1aa36dcc58019f5b8065c1ebcd1f1705c792d1ad885d64e"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import h5py\n",
      "import theanets\n",
      "import time"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Couldn't import dot_parser, loading of dot files will not be possible.\n"
       ]
      }
     ],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "f=h5py.File(\"data/data.hdf5\")\n",
      "f.keys()\n",
      "f['data'].keys()\n",
      "input_matrix = f['data']['input'].value\n",
      "target_matrix = f['data']['target'].value\n",
      "decorator_matrix = f['data']['deco'].value\n",
      "f.close()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "## take only first .. samples\n",
      "truncate=10000\n",
      "\n",
      "nInputs = input_matrix.shape[1]\n",
      "nOutputs = target_matrix.shape[1]\n",
      "\n",
      "input_matrix = input_matrix[0:truncate,...]\n",
      "target_matrix = target_matrix[0:truncate,...]\n",
      "decorator_matrix = decorator_matrix[0:truncate,...]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "(10000, 82)\n"
       ]
      }
     ],
     "prompt_number": 5
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "input_matrix.shape"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 7,
       "text": [
        "(10000, 82)"
       ]
      }
     ],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def allincluded( truncated, nHiddenLayers, epochs, trainings):\n",
      "    timing = 0\n",
      "    fom = 1.0\n",
      "    \n",
      "    exp=theanets.Experiment(theanets.feedforward.Regressor(layers=(nInputs,)+tuple([nInputs for l in range(nHiddenLayers)])+ ( (nOutputs, 'sigmoid'),)),'layerwise')    \n",
      "    \n",
      "    \n",
      "    cut   =int(input_matrix.shape[0]*0.8)\n",
      "    train =input_matrix[:cut], target_matrix[:cut]\n",
      "    valid =input_matrix[cut:], target_matrix[cut:]\n",
      "    \n",
      "    return (fom, timing)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 8
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "nInputs = input_matrix.shape[1]\n",
      "nOutputs = target_matrix.shape[1]\n",
      "nHiddenLayers = 3\n",
      "exp=theanets.Experiment(theanets.feedforward.Regressor(layers=(nInputs,)+tuple([nInputs for l in range(nHiddenLayers)])+ ( (nOutputs, 'sigmoid'),)),'layerwise')\n",
      "\n",
      "cut=int(input_matrix.shape[0]*0.8)\n",
      "train=input_matrix[:cut], target_matrix[:cut]\n",
      "valid=input_matrix[cut:], target_matrix[cut:]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 9
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "nEpoch=10\n",
      "nTrain=2\n",
      "for i in range(0,nEpoch):\n",
      "    print 'Learning rate=', 0.1/10**i\n",
      "    print 'Epoch',i+1,'of',nEpoch\n",
      "    for j in range(nTrain):\n",
      "        Start=time.mktime(time.gmtime())\n",
      "        print 'Iteration No.',j+1,'of',nTrain\n",
      "        print exp.train(train, valid, learning_rate=0.1/10**i, batch_size=100)        \n",
      "        End=time.mktime(time.gmtime())\n",
      "        print 'Time=',End-Start\n",
      "    print ''"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Learning rate= 0.1\n",
        "Epoch 1 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.39662499999999995), ('err', 0.39662499999999995)]), OrderedDict([('loss', 0.39750000000000008), ('err', 0.39750000000000008)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.39662499999999989), ('err', 0.39662499999999989)]), OrderedDict([('loss', 0.39750000000000002), ('err', 0.39750000000000002)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 0.01\n",
        "Epoch 2 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.016570155160109713), ('err', 0.016570155160109713)]), OrderedDict([('loss', 0.17746649049038299), ('err', 0.17746649049038299)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 26.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.018712363487130103), ('err', 0.018712363487130103)]), OrderedDict([('loss', 0.18241561169795686), ('err', 0.18241561169795686)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 0.001\n",
        "Epoch 3 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.01417446705078928), ('err', 0.01417446705078928)]), OrderedDict([('loss', 0.15045599070644994), ('err', 0.15045599070644994)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 27.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.0139340845808975), ('err', 0.0139340845808975)]), OrderedDict([('loss', 0.15290373239175942), ('err', 0.15290373239175942)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 24.0\n",
        "\n",
        "Learning rate= 0.0001\n",
        "Epoch 4 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.028346650884179814), ('err', 0.028346650884179814)]), OrderedDict([('loss', 0.12766284634614816), ('err', 0.12766284634614816)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.028350240054832138), ('err', 0.028350240054832138)]), OrderedDict([('loss', 0.12744380222402563), ('err', 0.12744380222402563)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 24.0\n",
        "\n",
        "Learning rate= 1e-05\n",
        "Epoch 5 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.035366412421076229), ('err', 0.035366412421076229)]), OrderedDict([('loss', 0.12295070575456424), ('err', 0.12295070575456424)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 24.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.03536871221528412), ('err', 0.03536871221528412)]), OrderedDict([('loss', 0.12295150197151976), ('err', 0.12295150197151976)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 1e-06\n",
        "Epoch 6 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.036376512415238919), ('err', 0.036376512415238919)]), OrderedDict([('loss', 0.12247285872657607), ('err', 0.12247285872657607)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 24.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.036376236442038738), ('err', 0.036376236442038738)]), OrderedDict([('loss', 0.12247392967097345), ('err', 0.12247392967097345)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 1e-07\n",
        "Epoch 7 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.03653862840255985), ('err', 0.03653862840255985)]), OrderedDict([('loss', 0.12240208258093635), ('err', 0.12240208258093635)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.036538571418122848), ('err', 0.036538571418122848)]), OrderedDict([('loss', 0.12240231558736692), ('err', 0.12240231558736692)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 1e-08\n",
        "Epoch 8 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.036559489488311933), ('err', 0.036559489488311933)]), OrderedDict([('loss', 0.12238541404365914), ('err', 0.12238541404365914)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.036559492867859512), ('err', 0.036559492867859512)]), OrderedDict([('loss', 0.12238540242214151), ('err', 0.12238540242214151)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 24.0\n",
        "\n",
        "Learning rate= 1e-09\n",
        "Epoch 9 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.03656170671797785), ('err', 0.03656170671797785)]), OrderedDict([('loss', 0.12238352422544255), ('err', 0.12238352422544255)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.036561707155270816), ('err', 0.036561707155270816)]), OrderedDict([('loss', 0.12238351508143035), ('err', 0.12238351508143035)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n",
        "Learning rate= 1e-10\n",
        "Epoch 10 of 10\n",
        "Iteration No. 1 of 2\n",
        "(OrderedDict([('loss', 0.036561929372197335), ('err', 0.036561929372197335)]), OrderedDict([('loss', 0.12238332815985309), ('err', 0.12238332815985309)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "Iteration No. 2 of 2\n",
        "(OrderedDict([('loss', 0.036561929386459538), ('err', 0.036561929386459538)]), OrderedDict([('loss', 0.1223833280855455), ('err', 0.1223833280855455)]))"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Time= 23.0\n",
        "\n"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "prediction = exp.network.predict(train[0])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 11
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.hist(prediction)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 12,
       "text": [
        "(array([ 2407.,   356.,   170.,   115.,    62.,    85.,    95.,   209.,\n",
        "          371.,  4130.]),\n",
        " array([  1.22707869e-04,   9.96970381e-02,   1.99271368e-01,\n",
        "          2.98845699e-01,   3.98420029e-01,   4.97994359e-01,\n",
        "          5.97568689e-01,   6.97143019e-01,   7.96717350e-01,\n",
        "          8.96291680e-01,   9.95866010e-01]),\n",
        " <a list of 10 Patch objects>)"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEACAYAAAC6d6FnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAFvVJREFUeJzt3X2QZXV95/H3B4YHdUYtBFx3elB2x2wM8jBYCiEam1Qw\nk8GoJVlMCYFCU7JuyFq7UVndVXpIURvXKklRWoYqFeMQhF2JRJcByqcpH0Ak0hRjj9Eqo9b0GGeA\nWFnZKCHDd/+4v565tEP3ne7pe3v6vF9Vp/qc7z3n3N891X0+9/zOOX1SVUiSuueIUTdAkjQaBoAk\ndZQBIEkdZQBIUkcZAJLUUQaAJHXUQAGQ5Mgkk0k+26Y/nuTvWm0yyemtniTXJZlKcn+SDX3ruLTV\np5JcsjQfR5I0qFUDzvc2YAewpk0X8Paq+qtZ870eOKmqTmk7/xuAM5I8D3gPcEab74Ekd1XV7sU1\nX5K0UPMeASQZAzYBHwHS/9IBZt8EbAGoqklgVVv+POCOqnq0qh4F7mw1SdKIDNIFdC3wDuCJWfVr\nknw7yQeTHNNqY8DOvnmmW21tG59dlySNyJwBkOTVwJ72bb7/G/+VVfXLwOnA0+h17+xb7JC3UpJ0\nyM13DuAc4DVJNgHHAs9M8omqugSgqv45yUeBq9r808A64N42PXNEMA2c1bfedcDds98sif+YSJIW\noKoO+sv3nEcAVfXuqlpXVScDvwd8saouSXIi9K76oXfid6otshW4qL12JrC3qnYBXwA2JlmTZA2w\nEfj8U7ynQxVXXXXVyNuwXAa3hdvCbTH3sFCDXgUEva6dmXe6Oclx9Lp/JoG3tJ33rUnOTTIFPAZc\n1uo/SnIN+48Mri6vAJKkkRo4AKpqG7Ctjf/GHPNd8RT1G+hdFipJWga8E3iZGh8fH3UTlg23xX5u\ni/3cFouXxfQfHWpJajm1R5IOB0moQ30SWJK0chkAktRRBoAkdZQBIEkdZQBIUkcZAJLUUQaAJHWU\nASBJHWUASFJHGQCS1FEGgCR1lAEgSR11MM8DkCQ1vedhHd4MAElasOXw34sXHkQDdQElOTLJZJLP\ntumTk9yTZHuSm5Mc1erHJLml1b+W5Pl963hXkh3ttVctuMWSpENi0HMAbwN2sD/urgPeV1WnAj8G\nZp4CdgXw963+/jYfSV5C79nBp9J7HvD1SY4+JJ9AkrQg8wZAkjFgE/CR3mSOBM6uqtvaLDcC57fx\nTcCWNv4Z4JwkR7TXb66qmYfETwEvO3QfQ5J0sAY5ArgWeAfwRJs+EXi47/VdwFgbHwN2AlTVE8Aj\nbf61wHTfMtN9y0iSRmDOk8BJXg3sqarJJOMz5aVs0MTExL7x8fFxn/spSb9gWxsWZ76rgM4BXpNk\nE3As8EzgfcDxffOMsf/b/TRwErCndf08B3io1dfNWmbngd6wPwAkSQcy3gaAzQtey5xdQFX17qpa\nV1UnA78HfLGqfh/4epLXtdkuBra28a1tGuC1wD1VtbfV35BkVTun8GLgGwtutSRp0Q72PoCZq4D+\nE3BTkj+hd0L37a3+QWBLku3AT4E3AlTVN5N8GniQ3rmEy6vq8cU2XpK0cKlaDjcy9CSp5dQeSXoq\nvTuBl8P+qndatqoO+vys/wtIkjrKAJCkjjIAJKmjDABJ6igDQJI6ygCQpI4yACSpowwASeooA0CS\nOsoAkKSOMgAkqaMMAEnqKANAkjrKAJCkjjIAJKmjDABJ6qg5AyDJsUnuSzKZ5LtJrm31jyf5u1af\nTHJ6qyfJdUmmktyfZEPfui5t9akklyztx5IkzWfOR0JW1c+T/HpV/SzJKuCrSc6l9xict1fVX81a\n5PXASVV1Stv53wCckeR5wHuAM9p8DyS5q6p2H9qPI0ka1LxdQFX1szZ6NHAksKdNH+jxY5uALW25\nSWDmIfDnAXdU1aNV9ShwZ6tJkkZk3gBIckSSB4DdwJeqaqq9dE2Sbyf5YJJjWm0M2Nm3+HSrrW3j\ns+uSpBGZswsIoKqeoNeN8yzgriTjwJVVtSfJ0cCH6XXv/Pe2yEE/mLjfxMTEvvHx8XHGx8cXszpJ\nWoG2tWFxUjX4U+2TvAd4vKr+tK92DnBVVf1Wko/S6+r5VHvtW8BvAb8BnFVVV7T6h4C7q+ovZ62/\nDqY9kjQqSeidDh213nfuqjroL9/zXQX0nCRr2vjT6PXbb09yYquF3onfmW6hrcBF7bUzgb1VtQv4\nArAxyZq2vo3A5w+2sZKkQ2e+LqB/DXyi7eiPBW6qqtuTfDHJccDTgEngLQBVdWuSc5NMAY8Bl7X6\nj5JcA9zb1nu1VwBJ0mgdVBfQUrMLSNLhYsV3AUmSVi4DQJI6ygCQpI4yACSpowwASeooA0CSOsoA\nkKSOMgAkqaMMAEnqKANAkjrKAJCkjjIAJKmjDABJ6igDQJI6ygCQpI4yACSpo+YNgCTHJrkvyWSS\n7ya5ttVPTnJPku1Jbk5yVKsfk+SWVv9akuf3retdSXa01161dB9LkjSfeQOgqn4O/HpVbQB+BfjV\nJOcC1wHvq6pTgR8DV7RFrgD+vtXf3+YjyUvoPT/4VHrPBL4+ydGH+PNIkgY0UBdQVf2sjR4NHAns\nAc6uqtta/Ubg/Da+CdjSxj8DnJPkiPb6zVU186D4KeBli/8IkqSFGCgAkhyR5AFgN/Al4CfAw32z\n7ALG2vgYsBOgqp4AHgFOBNYC033LTPctI0kaslWDzNR25GckeRZwF/DAUjVoYmJi3/j4+Djj4+NL\n9VaSdJja1obFGSgAZlTVPya5Hfg3wPF9L42x/9v9NHASsKd1/TwHeKjV181aZufs9+gPAEnSgYy3\nAWDzgtcyyFVAz0mypo0/DTiP3hHA15O8rs12MbC1jW9t0wCvBe6pqr2t/oYkq5KMAS8GvrHglkuS\nFiVVNfcMyanAJ4AAxwI3VdXVSU4GbgJW0zuh+/tV9XiSY+idBH4R8FPgjVX1g7aud9MLhyeAP66q\nu2a9V83XHklaDpIAy2F/FQCqKge95HLa4RoAkg4XKyEAvBNYkjrKAJCkjjIAJKmjDABJ6igDQJI6\nygCQpI4yACSpowwASeooA0CSOsoAkKSOMgAkqaMMAEnqKANAkjrKAJCkjjIAJKmj5gyAJOuSfDnJ\n9iTfSfLOVp9IMp1ksg2/3bfMu5LsaMu8qq++sdV2JLly6T6SJGkQcz4QJslzgROq6ltJVgP3A/8e\neB3w06r6wKz5XwL8OXA28K+ArwK/RC9o/hZ4ObAbuAd4S1VNzlreB8JIOiyshAfCzPlQ+KraTW+H\nTVU9muRBYO2T3vXJzgdubs8A3pVkCjiLXgBMVdUugCS3tHknD7AOSdIQDHwOIMkLgJcCX2mlP0zy\n7SQ3Jjmu1dYC032LTQNjrb7zAHVJ0ogMFACt++d/A2+rqp8CHwL+LfArwPeA65ashZKkJTFnFxBA\nkqOAW4Gbquo2gKp6uO/164EvtclpYF3f4mP0vvkfMau+jicfEewzMTGxb3x8fJzx8fH5P4Ukdcq2\nNizOfCeBA/wF8EhV/ee++olVtaeN/xFwblW9vu8k8K+y/yTwC4Ej6Z0E/jVgD3A3cHlV3T/r/TwJ\nLOmwsOJPAtPbYV8MPJhk5oTtu4E3JjkNOBr4IfDm1oBvJvk08CDwBL2d/OPA40neCtxF72hgy+yd\nvyRpuOY8Ahi2JLV79+5RN4PjjjuOVavm7R2T1GEr4Qhg2QXAsceeMNI2PPbYI3z3u99h/fr1I22H\npOVtJQTAsvua+/Of7xnp+69e7Y5fUjf4v4AkqaMMAEnqKANAkjrKAJCkjjIAJKmjDABJ6igDQJI6\nygCQpI4yACSpowwASeooA0CSOsoAkKSOMgAkqaMMAEnqqHkDIMm6JF9Osj3Jd5K8s9WPS/K5JA8m\nuSvJs/uWuS7JVJL7k2zoq1/a6lNJLlmajyRJGsQgRwD/DPzHqjoVeAnwB0lOBzYDt1fVacAdbZok\nFwAnVdUp9B4VeUOrPw94D3BWG96b5LmH+PNIkgY0bwBU1e6q+lYbf5Te837XApuALW22G4Hz2/j5\nM/WqmgRWJRkDzgPuqKpH23rubDVJ0ggc1DmAJC8AXgp8FTihqh4BqKqHgRPbbGuBnX2LTQNjrT59\ngLokaQQGDoAkq4FPAW+rqv873+yLapUkackN9EzgJEcBtwJ/WVW3tfJDSY6vqoeTnADMPMx3GlgH\n3Numx+gdEUzT6/ufsQ64+xffbaJvfLwNkqT9trVhceYNgCQBPgrsqKpr+17aClwM/Fn7uXVW/VNJ\nzgT2VtWuJF8Arkqyps23Ebj6F99xYkEfRJK6Y5z9X443L3gtgxwB/Bq9HfqDSSZb7V3AVcAtSd4E\n/Bi4EKCqbk1ybpIp4DHgslb/UZJr2H9kcHVV7V5wyyVJi5KqGnUb9klSMNr2rF69nsnJO1m/fv1I\n2yFpeet1jiyH/WfvlGtVHfS5V+8ElqSOMgAkqaMMAEnqKANAkjrKAJCkjjIAJKmjDABJ6igDQJI6\nygCQpI4yACSpowwASeooA0CSOsoAkKSOMgAkqaMMAEnqKANAkjpq3gBI8rEku5Ns76tNJJlOMtmG\n3+577V1JdiTZnuRVffWNrbYjyZWH/qNIkg7GIEcAN9B7fm+/Aj5QVRvacAdAkpcArwdObctcn+So\nJMcAH26104DfTbLhUH0ISdLBmzcAquorwE8O8NKBHj92PnBzVe2tql3AFHBWG6aqaldV/QtwS5tX\nkjQiizkH8IdJvp3kxiTHtdpaYLpvnmlgrNV3HqAuSRqRVQtc7kPA1W18ArgOuPhQNKi3uhnjbZAk\n7betDYuzoACoqodnxpNcD3ypTU4D6/pmHaP3zf+IWfV1PPmIoM/EQpokSR0yzv4vx5sXvJYFdQEl\nObFv8gJ6ff0AW4E3JFmVZAx4MfAN4D7gxUnWJjkKuBC4Y8GtliQt2rxHAEk+CbwSOD7JTuAq4Nwk\npwFHAz8E3gxQVd9M8mngQeAJ4PKqehx4PMlbgbvohc6Wqrp/KT6QJGkwqapRt2GfJNW7wnR0Vq9e\nz+Tknaxfv36k7ZC0vCVh1Purnt4FmVV1oCsz5+SdwJLUUQaAJHWUASBJHWUASFJHGQCS1FEGgCR1\nlAEgSR1lAEhSRxkAktRRBoAkdZQBIEkdZQBIUkcZAJLUUQaAJHWUASBJHWUASFJHzRsAST6WZHeS\n7X2145J8LsmDSe5K8uy+165LMpXk/iQb+uqXtvpUkksO/UeRJB2MQY4AbgA2zqptBm6vqtPoPdt3\nM0CSC4CTquoUeo+JvKHVnwe8BzirDe9N8txD8gkkSQsybwBU1VeAn8wqbwK2tPEbgfPb+Pkz9aqa\nBGYeDn8ecEdVPVpVjwJ3tpokaUQWeg7ghKp6BKCqHgZObPW1wM6++aaBsVafPkBdkjQiq5ZgnQf9\nYOInm+gbH2+DJGm/bW1YnIUGwENJjq+qh5OcAOxp9WlgHXBvmx6jd0QwTa/vf8Y64O4Dr3pigU2S\npK4YZ/+X480LXstCu4C2Ahe38Yvb9Ez9IoAkZwJ7q2oX8AVgY5I1SdbQO6n8+QW3WpK0aPMeAST5\nJPBK4PgkO4H3AlcBtyR5E/Bj4EKAqro1yblJpoDHgMta/UdJrmH/kcHVVbX7kH8aSdLAUlWjbsM+\nSQpG257Vq9czOXkn69evH2k7JC1vSRj1/qqnd9q1qg76/Kt3AktSRxkAktRRBoAkdZQBIEkdZQBI\nUkcZAJLUUQaAJHWUASBJHWUASFJHGQCS1FEGgCR1lAEgSR1lAEhSRxkAktRRBoAkdZQBIEkdtagA\nSPKDJA8mmUzyjVY7LsnnWv2uJM/um/+6JFNJ7k+yYbGNlyQt3GKPAAoYr6oNVfWyVtsM3F5VpwF3\ntGmSXACcVFWnAG8Gbljke0uSFuFQdAHNfgzZJmBLG78ROL+Nnz9Tr6pJYFWSsUPw/pKkBTgURwAz\n3T1XtNoJVfUIQFU9DJzY6muBnX3LTgMGgCSNyKpFLn92Ve1JcgJwZ5K/nWf+2UcLB3ii8kTf+Hgb\nJEn7bWvD4iwqAKpqT/v5UJJPAS8FHkpyfFU93IJhT5t9GlgH3Numx1ptlonFNEmSOmCc/V+ONy94\nLQvuAkry9CRPb+PPADYCU8BW4OI228Vtmvbzojb/mcDeqtq10PeXJC3OYo4AngvclqSApwM3V9Vn\nknwVuCXJm4AfAxcCVNWtSc5NMgU8Bly2yLZLkhZhwQFQVd8HTj9A/R+A855imSsOVJckDZ93AktS\nRy32KiBJGrpk9gWFWggD4ABe+MIXjroJAFQd4CpZSc2o/z4O/xAyAJ6Sv1ySVjbPAUhSRxkAktRR\nBoAkdZQBIEkdZQBIUkd5FZCkg+I1+CuHASBpAbxMeiUwAJax5fBNy5vRpJXLAFjWRr3zHX0ASVo6\nngSWpI7yCEBzWg7dUGBXlLQUhhoASTYC7weOBP6iqt43zPfXQiyHHe/yCKFRWy5hrJVjaF1ASY4B\nPkzv0ZGnAb+bZMOw3l9ajCQjH3q+RC+URzksF9tG3YDD3jDPAZwFTFXVrqr6F+AW4Pwhvr8OY8tj\n57scdrzblmoTH4a2jboBh71hBsAYsLNverrVpAEsh52vtLIM8xzAQH9Fz3zm7yx1O+b0T//0o5G+\nvyQNS4Z1dUWSVwBXVtWr2/Q7gKOr6pq+efyqJUkLUFUHfZXAMI8A7gNenGQtsAe4ELi8f4aFfABJ\n0sIMLQCq6udJ3grcRe/cw5aqun9Y7y9JerKhdQFJkpaXkfwriCQbk2xPsiPJlQd4/Zgkt7R5vpbk\n+aNo5zAMsC3ekWQqybeSfDnJyaNo5zDMty365rsgyRNJzhxm+4ZpkG2R5MIkk0keTHLTsNs4LAP8\njfxyknvb38iOJK8dRTuXWpKPJdmdZPsc81zX9hf3D3SfVVUNdQCOAb4PrKXXBXUfsGHWPH8M/Fkb\nfx3w18Nu5zLaFq8Ajmnj/wH49KjbPapt0eZbA3wZuBs4c9TtHuHvxenAvcAz2vRxo273CLfFjcDl\nbfxFwM5Rt3uJtsUrgA3A9qd4/QLgtja+AXhgvnWO4ghgkBvCNgFb2vhngHOyMu+Dn3dbVNVXquqx\nNvk1en8IK9GgNwr+CfCnwGOs3P8RMci2uAz4YFX9P4Cq+ocht3FYBtkWO4FntfFnAz8cYvuGpqq+\nAvxkjln27TerahJYlWTOe61GEQCD3BC2b56qegJ4BDhxKK0broO9Oe5y4K+XtEWjM++2aF0+a6tq\nayut1BNYg/xe/DvgjCR/k+SbSV4ztNYN1yDb4n8AlybZCdwO/NGQ2rbcHPTNtqP4b6Ar9Y92IQbe\nFkkuAs4EXrl0zRmpObdFkiOADwCX9peXtEWjM8jvxRHAC+h9Q14H3J3kqyvwSGCQbfEB4CNVdW2S\ns+l1CZ2ytM1atmb/Tcy5/UZxBDBN7xd2xjqenFoz85wE+/7wnwM8NJTWDdcg24Ikvwn8N+A1VfX4\nkNo2bPNtizX0/qi3Jfk+cDbwmRV6IniQ34udwGeram9V/QDYAfzScJo3VINsi5cD/wugqr4OHJtk\nJfYYzGf2thprtac0igDYd0NYkqPo3RB2x6x5tgIXt/HXAve0rqCVZt5t0c7k/znwO1X18AjaOCxz\nbouq+seqOqGqTq6qk4Gv09smK/FekkH+Rm4HxgGSHE/v5Of3htnIIRlkW3wP+E2AJC8CnkGv27hr\ntgIXwb7u0r1VtWuuBYbeBVRPcUNYks3A31TVZ4EPAlva5U4/Bd447HYOwzzb4r6q+j/A/6T3C/2p\ndh78h1X1upE1eokM+HvRCYNsi6r6dJKXJ5mi93yN/1pVK+4oecDfi/8CfDzJO+l1gfxBVe0dXauX\nRpJP0usCPr6d77gKOAqgqq6vqluTnNt+Jx6jd6HA3OtslwxJkjrGZwJLUkcZAJLUUQaAJHWUASBJ\nHWUASFJHGQCS1FEGgCR1lAEgSR31/wFuw73AwDW13gAAAABJRU5ErkJggg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7fc99155c8d0>"
       ]
      }
     ],
     "prompt_number": 12
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.hist(target_matrix)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 13,
       "text": [
        "(array([ 3968.,     0.,     0.,     0.,     0.,     0.,     0.,     0.,\n",
        "            0.,  6032.]),\n",
        " array([ 0. ,  0.1,  0.2,  0.3,  0.4,  0.5,  0.6,  0.7,  0.8,  0.9,  1. ]),\n",
        " <a list of 10 Patch objects>)"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEACAYAAAC6d6FnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7fc979debd90>"
       ]
      }
     ],
     "prompt_number": 13
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "goodness = np.asarray([[cut,(list(target_matrix[np.where(prediction>cut)]==1).count(True)+list(target_matrix[np.where(prediction<cut)]==0).count(True))/ float(input_matrix.shape[0])]  for cut in np.arange(0,1,0.05)])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 14
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.scatter(goodness[:,0], goodness[:,1])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 15,
       "text": [
        "<matplotlib.collections.PathCollection at 0x7fc979d43150>"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEACAYAAABfxaZOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ny8TO+wAAAABJRU5ErkJggg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7fc97927d410>"
       ]
      }
     ],
     "prompt_number": 15
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "figure_of_merit = max( goodness[:,1])\n",
      "print figure_of_merit"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.7692\n"
       ]
      }
     ],
     "prompt_number": 16
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "exp.network.layers[4].params[0].get_value().shape"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 17,
       "text": [
        "(82, 1)"
       ]
      }
     ],
     "prompt_number": 17
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "exp.network.num_params"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 18,
       "text": [
        "20501"
       ]
      }
     ],
     "prompt_number": 18
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}