500K4L-relu-rprop.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:0fcc4e9be573fce2956f5895f5358f6972482a12970ba88fab09101250f92d3c"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import theanets\n",
      "import logging\n",
      "from sklearn.datasets import make_classification\n",
      "from sklearn.metrics import confusion_matrix\n",
      "import matplotlib.pyplot as plt\n",
      "import time\n",
      "import numpy as np\n",
      "from theanets.trainer import SupervisedPretrainer\n",
      "import math\n",
      "from random import uniform\n",
      "from math import cos,tan,sqrt\n",
      "from numpy import float32\n",
      "from numpy import int32\n",
      "from sklearn.cluster import KMeans\n",
      "import pickle"
     ],
     "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": [
      "nlayers = 4\n",
      "module_size=20 #cm\n",
      "module_cell = 5 #per edge\n",
      "distance_between_layers = 5 #cm\n",
      "ncells=module_cell*module_cell*nlayers\n",
      "nhidden=ncells\n",
      "nhidden_layers = 10\n",
      "layers = (ncells,) + tuple([nhidden for i in range(nhidden_layers)])\n",
      "#net = theanets.Regressor((ncells,ncells,ncells,ncells,ncells,ncells,ncells,ncells,ncells,ncells,ncells,(ncells,'logistic')))\n",
      "trainer = theanets.Experiment(theanets.feedforward.Regressor(layers=(ncells,(ncells,'relu'),(ncells,'relu'),(ncells,'relu'),(ncells,'relu'),(ncells,'logistic'))),'layerwise')\n",
      "maximum_hight = nlayers * distance_between_layers\n",
      "theta_max = math.atan(0.5 * module_size / maximum_hight)\n",
      "theta_min = -1 * theta_max\n",
      "y = []\n",
      "y.append(0)\n",
      "y_prime = []\n",
      "y_prime.append(0)\n",
      "z = []\n",
      "z.append(0)\n",
      "x = []\n",
      "x.append(0)\n",
      "for i in range (nlayers) :\n",
      "    y.append((i+1) * distance_between_layers)\n",
      "    y_prime.append((i+1) * distance_between_layers)\n",
      "    z.append(0)\n",
      "    x.append(0)\n",
      "spent = []\n",
      "list_error = []\n",
      "color_list = ['bo','yo','ro','co','mo','go','ko','wo']\n",
      "min_ntracks = 4\n",
      "max_ntracks = 4\n",
      "nsamples = 500000\n",
      "#nsamples = 2 * ncells * ncells * (nhidden_layers + 1)\n",
      "#print nsamples\n",
      "training_ratio = 0.8\n",
      "ds_input = np.zeros((nsamples,ncells))\n",
      "ds_output = np.zeros((nsamples,ncells))\n",
      "cut = int(nsamples * training_ratio)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def addSample (sample_number = 1,sampleInput = np.zeros (ncells),sampleOutput = np.zeros (ncells)) :\n",
      "    for i in range (ncells) :\n",
      "        ds_input[sample_number][i] = sampleInput[i]\n",
      "        ds_output[sample_number][i] = sampleOutput[i]\n",
      "    \n",
      "def sample_creater(sample_number = 1,display = False) :                  # creates a sample with a random number of tracks\n",
      "    sampleInput = np.zeros (ncells)\n",
      "    sampleOutput = np.zeros (ncells)\n",
      "    ntracks = int( uniform(min_ntracks, max_ntracks ) ) #should be a better model than flat\n",
      "    sample_matrix = [[0 for t in range(nlayers + 1)] for t in range(ntracks)]\n",
      "# first argument of each row is track's radius and the other \"nlayers\" ones are hit indices of that track\n",
      "# then we would be able to sort these hits by their radius\n",
      "    \n",
      "# theta is the angle between the circle plane and xy plane. we randomly choose two points in the second and 4th layer and then fit a circle through these two points and also the coordinate origin\n",
      "# as y is fixed for each layer and z can be derived easily by y*tan(theta) , to specify a point in a layer we only need x of it\n",
      "    track = 0\n",
      "    while track < ntracks :\n",
      "        theta = uniform (theta_min,theta_max)\n",
      "        x[2] = uniform (-1*module_size/2.0,module_size/2.0)\n",
      "        x[4] = uniform (-1*module_size/2.0,module_size/2.0)\n",
      "        for i in range (4) :\n",
      "            y_prime[i+1] = y[i+1] / cos(theta)\n",
      "            z[i+1] = y[i+1] * tan(theta)\n",
      "        # center point of our circle\n",
      "        x_c = 0.5 * (((x[2]**2 + y_prime[2]**2) * y_prime[4]/y_prime[2] ) - x[4]**2 - y_prime[4]**2 )/ (x[2] * y_prime[4]/y_prime[2] - x[4])\n",
      "        y_c = 0.5 * (x[2]**2 + y_prime[2]**2 - 2 * x[2] * x_c) / y_prime[2]\n",
      "        R = sqrt(x_c**2 + y_c**2)\n",
      "# y_prime is the y component when we place our coordinate on the circle plane and as the layers are parallel, it would be y/cos(theta) for each layer\n",
      "        if x[2] > 0.5 * x[4] :\n",
      "            x[1] = x_c + sqrt(R**2 - (y_prime[1] - y_c)**2)\n",
      "            x[3] = x_c + sqrt(R**2 - (y_prime[3] - y_c)**2)\n",
      "        else :\n",
      "            x[1] = x_c - sqrt(R**2 - (y_prime[1] - y_c)**2)\n",
      "            x[3] = x_c - sqrt(R**2 - (y_prime[3] - y_c)**2)\n",
      "        sample_matrix [track][0] = R\n",
      "        x_edge = (-1) * module_size / 2.0\n",
      "        z_edge = (-1) * module_size / 2.0\n",
      "        \n",
      "        track_hits = np.zeros(nlayers)\n",
      "        no_good = False\n",
      "        for layer in range (1,nlayers+1) :\n",
      "            # local indices on the module\n",
      "            index_hit1 = int( (x[layer] - x_edge ) / module_size * module_cell )\n",
      "            index_hit2 = int( (z[layer] - z_edge ) / module_size * module_cell )\n",
      "            # transform into the global index\n",
      "            hit = module_cell**2 * (layer - 1) + index_hit2 * module_cell + index_hit1\n",
      "            track_hits[layer-1] = hit\n",
      "            if sampleInput[hit] == 1:\n",
      "                    #discard this track\n",
      "                    no_good = True\n",
      "        if no_good:\n",
      "            #print \"discarding an overlaping track\"\n",
      "            continue\n",
      "        else:\n",
      "            for (layer,hit) in enumerate(track_hits):\n",
      "                sampleInput[int(hit)] = 1\n",
      "                sample_matrix [track][layer+1] = int(hit)\n",
      "        if display:\n",
      "            for i in range (0,5) :\n",
      "                plt.figure(1)\n",
      "                plt.plot(x[i],y_prime[i],color_list[track])\n",
      "                axis([-module_size/2. + 1 , module_size/2. + 1, 0, maximum_hight/cos(theta) + 1])\n",
      "                plt.suptitle('x - y_prime (perpendicular view of the circle plane)')\n",
      "                plt.xlabel('x')\n",
      "                plt.ylabel('y_prime')\n",
      "                plt.figure(2)\n",
      "                plt.plot(x[i],z[i],color_list[track])\n",
      "                axis([-module_size/2. + 1, module_size/2. + 1,-module_size/2. + 1, module_size/2. + 1])\n",
      "                plt.suptitle('x - z')\n",
      "                plt.xlabel('x')\n",
      "                plt.ylabel('z')\n",
      "                plt.figure(3)\n",
      "                plt.plot(z[i],y[i],color_list[track])\n",
      "                axis([-module_size/2. + 1, module_size/2. + 1, 0, maximum_hight/cos(theta) + 1])\n",
      "                plt.suptitle('z - y')\n",
      "                plt.xlabel('z')\n",
      "                plt.ylabel('y')\n",
      "        track += 1\n",
      "    sample_matrix.sort(key=lambda x: x[0])\n",
      "    for i in range (ntracks) :\n",
      "        for j in range(1,nlayers+1) :\n",
      "            sampleOutput[sample_matrix[i][j]] = (i+1)/float(ntracks)\n",
      "    if display == False :\n",
      "        addSample(sample_number,sampleInput,sampleOutput)\n",
      "        \n",
      "def new_samples() :\n",
      "    ds_input = np.zeros((nsamples,ncells))\n",
      "    ds_output = np.zeros((nsamples,ncells))\n",
      "    for i in range (nsamples) :\n",
      "        sample_creater(i,False)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 3
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def clusterer (in_array,revised_number = 0) :\n",
      "    n_lines = len(in_array)\n",
      "    n_gaps = n_lines - 1\n",
      "    gap_matrix = np.zeros((n_gaps,3),dtype=float32)\n",
      "    length_array = np.zeros(n_gaps,dtype=float32)\n",
      "    for i in range (n_gaps) :\n",
      "        gap_matrix[i][2] = i\n",
      "        gap_matrix[i][0] = in_array[i+1] - in_array[i]\n",
      "        length_array[i] = in_array[i+1] - in_array[i]\n",
      "        gap_matrix[i][1] = (in_array[i+1] + in_array[i])/2.0\n",
      "    length_array = sorted(length_array)\n",
      "    gap_matrix.view('f32,f32,f32').sort(order=['f0'], axis=0)\n",
      "    out_matrix = np.zeros((int(n_lines),2),dtype=int32)\n",
      "    if revised_number == 0 :\n",
      "        col = np.reshape(length_array,(len(length_array),1))\n",
      "        algo = KMeans(n_clusters=2)\n",
      "        gap_candidates = algo.fit_predict(col)\n",
      "        if gap_candidates[0] == 0:\n",
      "            accepted = list(gap_candidates).count(1)\n",
      "        else :\n",
      "            accepted = list(gap_candidates).count(0)\n",
      "    else :\n",
      "        accepted = revised_number-1\n",
      "    for i in range (n_gaps-accepted) :\n",
      "        gap_matrix[i][1] = in_array[n_gaps]\n",
      "    gap_matrix.view('f32,f32,f32').sort(order=['f1'], axis=0)\n",
      "    if accepted == 0 :\n",
      "        out_matrix [0] = [0,n_gaps]\n",
      "        return accepted+1,out_matrix\n",
      "    out_matrix [0] = [0,gap_matrix[0][2]]\n",
      "    out_matrix [accepted] = [gap_matrix[accepted-1][2]+1,n_gaps]\n",
      "    for bunch in range (1,accepted) :\n",
      "        out_matrix [bunch] = [int(gap_matrix[bunch-1][2]+1),int(gap_matrix[bunch][2])]\n",
      "    return accepted+1,out_matrix\n",
      "\n",
      "def clustering (in_array) :\n",
      "    central_tracks ,_ = clusterer (in_array)\n",
      "    col = np.reshape(in_array,(len(in_array),1))\n",
      "    min_n = max(1,int((0.75*central_tracks)))\n",
      "    max_n = int(1.25*central_tracks)\n",
      "    if max_n <= min_n + 1 :\n",
      "        return clusterer (in_array,max_n)\n",
      "    goc = np.zeros(1+max_n-min_n)\n",
      "    for nclu in range(min_n ,max_n +1):\n",
      "        algo = KMeans(n_clusters=nclu)\n",
      "        track_candidates = algo.fit_predict(col)\n",
      "        track_candidate_values = [np.array(in_array)[np.where(track_candidates==n)] for n in range(nclu)]\n",
      "        goc[nclu - min_n] = algo.score(col)\n",
      "    d_goc = []\n",
      "    for i in range (max_n-min_n):\n",
      "        d_goc.append(goc[i+1] - goc[i])\n",
      "    col = np.reshape(d_goc,(len(d_goc),1))\n",
      "    algo = KMeans(n_clusters=2)\n",
      "    gap_candidates = algo.fit_predict(col)\n",
      "    if gap_candidates[0] == 0:\n",
      "        accepted = list(gap_candidates).count(1)\n",
      "    else :\n",
      "        accepted = list(gap_candidates).count(0)\n",
      "    revised_number = max_n-accepted\n",
      "    return clusterer (in_array,revised_number)\n",
      "def similarity (exptrack,nntrack) :\n",
      "    sim_counter = 0\n",
      "    for exphit in range (len(exptrack)) :\n",
      "        for nnhit in range (len(nntrack)) :\n",
      "            if exptrack[exphit] == nntrack[nnhit] :\n",
      "                sim_counter += 1\n",
      "    return sim_counter\n",
      "def got (exp,nn,cut = 0.1,requested_efficiency = 1) : # Dear reader, get ready for awful variable names and no comment !\n",
      "    goodTrack = 0\n",
      "    # creating MidPut\n",
      "    for i in range (len(nn)) :\n",
      "        if nn[i] < cut :\n",
      "            nn[i] = 0\n",
      "    exp_Output = np.array([[i,exp[i]] for i in range (len(exp)) if exp[i] != 0],dtype = float32)\n",
      "    nn_Output = np.array([[i,nn[i]] for i in range (len(nn)) if nn[i] != 0],dtype = float32)\n",
      "    exp_Output.view('f32,f32').sort(order=['f1'], axis=0)\n",
      "    nn_Output.view('f32,f32').sort(order=['f1'], axis=0)\n",
      "    n_exp,exp_matrix = clustering (exp_Output[:,1])\n",
      "    n_nn,nn_matrix = clustering (nn_Output[:,1])\n",
      "    expM = np.array([[exp_Output[i][0] for i in range (exp_matrix[track][0],exp_matrix[track][1]+1)] for track in range (n_exp)])\n",
      "    nnM = np.array([[nn_Output[i][0] for i in range (nn_matrix[track][0],nn_matrix[track][1]+1)] for track in range (n_nn)])\n",
      "    similarities = np.zeros(n_exp,dtype = float32)\n",
      "    # doesn't work quite well for efficiencies less than 1,I know the reason,but don't have time to fix it !\n",
      "    remained_NNtracks = np.array([i for i in range (len(nnM))])\n",
      "    for exptrack in range (n_exp) :\n",
      "        for nntrack in remained_NNtracks :\n",
      "            if similarity(expM[exptrack],nnM[nntrack]) >= requested_efficiency * len(expM[1]) :\n",
      "                goodTrack +=1\n",
      "                remained_NNtracks = np.array([remained_NNtracks[i] for i in range (len(remained_NNtracks)) if i != nntrack]) #like !\n",
      "        #similarities[exptrack] = max([similarity(expM[exptrack],nnM[nntrack]) for nntrack in range (len(nnM))])\n",
      "    for exptrack in range (n_exp) :\n",
      "        if similarities[exptrack] >= requested_efficiency * len(expM[1]) :\n",
      "            goodTrack +=1\n",
      "    return float(goodTrack)/len(expM)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 4
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "new_samples()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 5
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "training_error = []\n",
      "validation_error = []\n",
      "train = ds_input[:cut], ds_output[:cut]\n",
      "valid = ds_input[cut:], ds_output[cut:]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 6
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# we try to create some training samples and train on them and then check the mean error in some testing samples\n",
      "# a function of iteration and train per iterations\n",
      "e = []\n",
      "for i in range(10) :\n",
      "    e.append(0)\n",
      "niter=1\n",
      "ntrain=1\n",
      "for i in range (1,niter+1) :\n",
      "    # training on training samples\n",
      "    for j in range (ntrain) :\n",
      "        start = time.mktime(time.gmtime())\n",
      "        print \"training started ... \"\n",
      "        epoch_counter = 0\n",
      "        for trn, val in trainer.itertrain(train,valid, algo='rprop',hidden_dropout=0.8,learning_rate = 0.000001,iteration_size = 100,batch_size=10000,validate_every = 1,patience=10,min_improvement = 0.001) :\n",
      "            if trn['err'] < 0.2 and val['err'] < 0.2 :\n",
      "                training_error.append(trn['err'])\n",
      "                validation_error.append(val['err'])\n",
      "                if epoch_counter%1 == 0 :\n",
      "                    print \"training_error = \\t\",trn['err']\n",
      "                    print \"validation_error = \\t\",val['err']\n",
      "                epoch_counter += 1\n",
      "        print \"\\t training ended\"\n",
      "        stop = time.mktime (time.gmtime())\n",
      "        spent.append (stop - start)\n",
      "        print \"training time (seconds) \\t\" , stop - start"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "training started ... \n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905957924741\n",
        "validation_error = \t0.00908323106692\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00906135353892\n",
        "validation_error = \t0.00908302012256\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905880404452\n",
        "validation_error = \t0.00908283383793\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00906113693172\n",
        "validation_error = \t0.00908262932331\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00906010802974\n",
        "validation_error = \t0.00908245747254\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905935624714\n",
        "validation_error = \t0.0090823240704\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905930530361\n",
        "validation_error = \t0.00908220402633\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905983388946\n",
        "validation_error = \t0.00908210309443\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00906058512135\n",
        "validation_error = \t0.00908227447998\n",
        "training_error = \t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.00905837391073\n",
        "validation_error = \t0.00908221599275\n",
        "\t training ended"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "training time (seconds) \t892.0\n"
       ]
      }
     ],
     "prompt_number": 8
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.plot(training_error,'b')\n",
      "print training_error[-1]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.0114509865828\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEACAYAAAC+gnFaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f537589f9d0>"
       ]
      }
     ],
     "prompt_number": 21
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.plot(validation_error,'b')\n",
      "print validation_error[-1]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.0113671492548\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEACAYAAAC+gnFaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f536fe5da50>"
       ]
      }
     ],
     "prompt_number": 22
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "e"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 11,
       "text": [
        "[(OrderedDict([('loss', 0.0085986440083445168), ('err', 0.0085986440083445168)]),\n",
        "  OrderedDict([('loss', 0.0086501840236289668), ('err', 0.0086501840236289668)])),\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0,\n",
        " 0]"
       ]
      }
     ],
     "prompt_number": 11
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "spent"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 10,
       "text": [
        "[5582.0]"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "trainer.save(\"network/500K4L-relu-rprop\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 14
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "trainer.load(\"network/500K4L-relu-rprop\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 7,
       "text": [
        "<theanets.feedforward.Regressor at 0x7faabacb0790>"
       ]
      }
     ],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.plot(trainer.network.predict(valid[0])[30])\n",
      "plt.xlabel('hit index')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 16,
       "text": [
        "<matplotlib.text.Text at 0x7f40ffbfd5d0>"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEPCAYAAABIut/fAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJvCGYRhdigm8YRhGl/L/AXdF7hyiuQcVAAAAAElFTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f40ffb73950>"
       ]
      }
     ],
     "prompt_number": 16
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.plot(ds_output[80004])\n",
      "plt.xlabel('hit index')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 10,
       "text": [
        "<matplotlib.text.Text at 0x7f92e0a858d0>"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEPCAYAAABIut/fAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXuQJWd53n/vzt60q8vKERCiXS5lzE0yIDkJ2EWswWCX\nbCFISkSJTaiEMlVEFBQuU0RJSNBu2RTBFyiDXDgxyK7ggOWAbMxFEbl4ipvkSEglpJUgBgPRCiIh\n4V3trjizc/nyR5/W6enpy9fdX/fp0/38qqZ2zpmeM73v6X7O28/3fF+bcw4hhBDDY8e8d0AIIUQ7\nSOCFEGKgSOCFEGKgSOCFEGKgSOCFEGKgSOCFEGKglAq8md1gZg+Z2T0F27zfzI6a2Z1mdknYXRRC\nCFEHnw7+D4DL835oZlcBT3POXQT88nR7IYQQc6ZU4J1zXwD+pmCTXwA+Mt32LmCnmR0Ms3tCCCHq\nEsKDPwg8kHh8bPqcEEKIORJqkNVSj7X+gRBCzJmdAV7jGHAI+Mvp44PT57ZgZhJ9IYSogXMu3UR7\nEaKD/yzwWgAzuxTYcM49mLWhc673X3ff7Th4sN2/cd111839/9mXL59avOxljptumv++9qEWTb9O\nn3aAY21t/v/fedei6Ot3f9dx1VXzr4Nzzfri0g7ezD4GXAZcYGYPANcBu6aC/R+dc58ws5eZ2VFg\nFXh9oz2aM+vrsLY2770QSdbXoy/RnLiO6+uwM8T1+0CZTIZxzJW+xc65X/TY5s1hdmf+SOD7hwQ+\nHEmBF/kMReA1kzVFFwK/vLzc7h9YIHxqMRaB7+K4WBSBn/c5srra/xr5IIFPIYHvFgn8DAn8jHmf\nI+rgB4osmv4xFoHvgkUR+HkjgR8oGxvRV8PBaxGQ+D0RzYnrqHoWs7o6jBpJ4FPEn9rq4vuDOvhw\nqIP3Qx38QInf1DNn5rsfYoYEPhwSeD8k8ANFHXz/kMCHQwLvh1I0A0UC3z8k8OGQwPuhDn6gSOD7\nhwQ+HBJ4PyTwAyUeOZfA9welaMKhFI0fStEMFHXw/UMdfDjUwfuhDn6gSOD7hwQ+HBJ4PyTwA0UC\n3z8k8OGQwPuhFM1AUQ6+f0jgwyGB92MoHbxWhE6hDr5fOBcNdg3hZOsDEng/JpNhLFeiDj6FUjT9\nYnMz+ncIiYY+oBSNH0rRDBR18P1CHWdYVE8/hmLRSOBTSOD7hQQpLKpnOevrw7EFJfApJPD9QoIU\nFtWznNVV2L17GDWSwKeQwPcLCVJYVM9yJhPYvz8a/4nHgBYVCXwKCXy/kCCFRfUsZzKBs86CpaXF\nH2iVwKeI31Dl4PuBUh9hiYVd9cxnMoG9e2HnzsWvkwQ+hTr4fqGOMyyxYKme+ayuwp49UQe/6HWS\nwKeQwPcLCXxYVM9ykh38otdJAp9ifR127JDA9wUJUlhUz3Ik8ANmfT0aYJHA9wMJUlhUz3Jii0YC\nP0Ak8P1CghQW1bMcdfADZn09enMl8P1AKZqwKEVTjlI0A2ZjQx18n1DHGRalaMpRimbAyKLpF+vr\nw7hU7guqZzmyaAZMLPCa6NQP1tejbmrRT7S+oHqWI4EfMOvrsG+fOvi+IEEKi+pZjlI0A0YWTb+I\nB70X/UTrC6pnOergB4xSNP1iYyN6PxY9zdAX4uNb9cxHKZoBoxRNv1DHGZb4A1P1zEcpmgEji6Zf\nSODDonqWMyqLxswuN7N7zOw+M7s24+fPNbO/NLN7p9u8up1d7QYJfL/QoGBYVM9yRiPwZrYH+CBw\nOfAC4DVmdklqs38H3OCcuxi4Cri+jR3tCgl8v5AghUX1LGdMKZoXA0edcw8659aBG4ErUts8AJw3\n/f4A8J2wu9gtysH3C1kKYVE9yxlSB7+z5OcHiQQ85hiwnNrm3cCtZvYWYD/w8mB7NwfUwfcLpWjC\nohRNOUNK0ZQJvPN4jfcCH3LOvc/MXgL8EXBR1oaHDx9+4vvl5WWWl5f99rJDlKLpF+o4wxI3MKpn\nPvO2aFZWVlhZWQnyWmUCfww4lHh8iK0dPcBLgesAnHO3mdleM3uyc+7h9IslBb6vqIPvFxL4sMRX\nRI89Nu896S9xBz+vmGS6+T1y5Ejt1yrz4G8HLjazC81sF3A1cHNqm28CrwAws+cR2TSP1t6jOSOB\n7xcaFAyL6lnOkDz4QoF3zk2Aa4BbgLuBm5xzd5rZETO7crrZrwL/0syOAp8A3uCcW1jnSgLfL9bX\nYdcucA42N+e9N4uPBL6cyWQ4KZoyiwbn3M2kunbn3HWJ778O/GT4XZsPEvh+EQt8fLLt3j3vPVps\nZHmVs7o6kg5+jEjg+0W8fvkQTrY+IIEvZzQWzRiJB6GUg+8HGxszgV/0yFofUEyynKRFs+h1ksCn\n0Hrw/UIdfFi02Fg5sUWjxcYGiCyafiGBD4ssmnJk0QwYCXy/WF+POqkhdFN9QCmacoaUohmVwPuI\ntgS+O9bWovhjEerg/fE9viXw+WxuztJabR5zXenLaAT+/vvhssvKt5PAd8eVV8IddxRvI4H34/Rp\neNazyreTRVNMvEyBWbvH3POeB8ePt/PaSUYj8MePw/e+V7zN5mbUUe7ZI4Hvgh/8oPwgV4rGj1On\n4KGHyrdTiqaY2J6Bdo+5hx+OPpTbZjQCv7bmLyZLSzOxF+0xmZR/kKqD98OnlqAUTRlxggbaHfdZ\nW+umiRyVwD/2WLFox2ISX56pi28XCXw4JpOoKSlbzkEWTTFxggba9+Al8AFZW4sO/lOn8reJxQSi\n6fGa7NQuq6t+Aq8UTTmrq9G/PvXUIGs+aYumjTo5F11JSeADEhezyKZJCvzu3erg20YdfDgmk+hf\nCXwzkhZNW8dc/B5J4AMSF/PEifxt4m4Rog5eAt8uEvhwVBF4WTT5dGHRSOBbwFfgkxaNBL5dfCwa\npWj8qGLRKEWTTxcpGgl8C/hYNLGYgAS+bZzz9+DVwZejDj4MXaRoJPAtEA+YqoPvB3HHWTaQLYH3\nIxb4snoqJllMFxZN/B51EeIYjcDLoukXVSwFpWjK8alnnN7YvVu1zKOLFI06+BaomqKRwLdLFUtB\nHXw5PvXc2Ig+KFXLfJSiWVDqdPDKwbeHBD4sPvWMa7ljh+5xm4dSNAvK2hrs3+8fk1QOvl18LRql\naPzwqWd6prbquZ20wCtFsyCsrcEFFyhF0xfUwYelSgcPqmce8WqSoA5+oVhbgyc9SYOsfUECHxZf\nD14CX0yyg1dMcoHw6eAl8N2hFE1YfC2a2IJUPbORB7+gqIPvF+rgwyKLJgxdxCTj8IYEPiBnzkjg\n+4TvxBwJvB8+9ZTAl9NlTFITnQLia9FosbFuUIomLFVSNKB65qEUzYKytgbnnw8//GH+p3I6RaMc\nfHtMJlFcTxZNGHzqqQ6+HM1kXVDW1qJs+7nnRnd2ykLrwXfHZAJnny2BD4VPPZWiKUeLjS0oa2tR\nV37eefk2jTz47lhdhXPOkcCHwqee6uDLUYpmQYkF/sCB/IFWCXx3VOngFZMsx6eeikmWI4tmQUl2\n8BL4+TOZqIMPiU891cGXo8XGFhRZNP3CV+CVovGjjsCrntvpIkVz5oxfwCAEoxH4M2eigdMiiyZe\nThUk8G2zuhpZCsrBh8Gnnurgy+nKotm3Tzn4oMii6ReyaMIiiyYMXaVo9u1TBx8UWTT9QgIfFp96\nKiZZTlcpmt4IvJldbmb3mNl9ZnZtzjZXm9ldZvZVM/to+N1sTp0UjSY6tUeVmKRSNOX4xiSVoimm\nS4umC4HfWfRDM9sDfBB4KfAQcKuZfc45d1dimxcCbwNe6pw7bWY/0uYO1yXZwd9/f/Y2mujUHT4d\nZ3zXId1mrpy4nidP5m8ji6acrlI0fengXwwcdc496JxbB24Erkht83rgeufcaQDn3A/C72ZzlIPv\nF76WwtKS7kBUhnP+OXilaIrpai2avgj8QeCBxONj0+eSPAd4kZndYWZfMbNXhdzBUPgMsipF0x1x\n6kODgs2JrZe9e1XPJjjX3R2demHRAM7jNXYAzyDq9g8BXzazL/atk9cga79Q6iMccddZdsyqnsWs\nrc1uSg7trge/b1/xyrahKBP4Y0SiHXOIrR0908dfdM5tAN82s/uAZwO3pV/s8OHDT3y/vLzM8vJy\n9T2uSbzYmCyafiCBD0c8MFh2zCpFU0zSnoF2Y5L798P3v5/985WVFVZWVoL8rTKBvx242MwuBB4G\nrgbemNrmM8CrgT80swuA5wHfzHqxpMB3zZkz0QmwZ0+xwMdvsAS+XXwn5ij1UU48MFiW/FI9i0km\naCDq5OOB/h0BA+WxwOe9V+nm98iRI7X/VuFuO+cmwDXALcDdwE3OuTvN7IiZXTnd5k+BR83sKPBF\n4F8753I+m+aHLJp+oQ4+HLJowpBM0MBscD90nfrkweOcuxm4OfXcdanHbyOKSvaWWODjNzB9OQbK\nwXfJZBKtzS9Bak58LJdFe1XPYrI0Ia7T7t3h/k6fUjSDIRZ4s/wkTfISVjn4dlldjS5TNzejryzS\nnrFifdnEyY+qHbzquZW0RQPt1EkC3wKxwEO+TZO+ZZ8Evj18bAV1nH7IoglDUQcfEgl8YJyL3qRY\n4POSNPLgu0MCHw5fgVeKppi0Bw/tDEZL4AMTWy9m0eMii0YC3z7JCSVlAq/URzlVLBrVM588iyZ0\nneIcvAQ+EHEGPibPopHAd8P6ehQ7W1pSBx8CWTRh6NKi2b9fAh+MOAMfI4tmviRPpKK0kgTJj6TA\n64Yf9cmyaNoUeN3wIxDJAVZQBz9vkut9FNVZKRo/lKIJg1I0C0pa4PM6+PRiY8rBt0Oygy+Ko6rj\n9EM5+DAoRbOgZHXwZRaNcvDtkbZoJPDNkAcfhi5TNGedFTWUzmc5xwaMVuBl0cwPX4tGguSHr0Wj\nmGQxXaVoYj3aubN9jRmlwGuQdb5U6eAV6yunSgeveubTpUWza1c3GjNKgVcOfr7IogmLLJowdJWi\niVN9EvhAyKLpF8lLYaVomuO7HrxSNMV0maLZvbubcb7RCHxyopNvikYC3w7JTkkdfHOS68Grg6+P\nLJoFJT3RyceiWVqKVjlse5R7jGiiU1g00SkMXU50igW+7Sj2KAQ+bdGce24k8GnxTp4A8WL/6uLD\nk7wUVg6+OXE9d+6stvyy6rmVrA6+rZikOviApAV+164oh3rq1NbtkoICkfhoslN4qlg0Sn2UE9fT\nTPVsQhcxSedmH7QS+ECkBR6ybZq0wMuHz+fRR+GRR+r9rlI0xRw/Dg895L+96hmGLiyatbXoNcs+\njEMxWoE/5xx47LGtz0ng/fnAB6KvOihFU8yHPwzveY//9r71VIqmmC5SNEkt6kJfSu/JOgSyBH7v\n3ugTO0kyRQMS+CIef7z+ALRSNMU8/nj05YvqGYYuUjQS+BbIE/jJZOtz6uD9WV2tL/B1LIXYM3Zu\nduOWobK6ur35KEIWTRi6sGiSiT5ZNIFI5+Ahu4OXwPszmWz/gKzyu1UthR07oq+8hMiQqFrbupaX\nBH4rXaRoklrUxUSn0Xbwe/aog2/CZNKNRZO0zOKTLfncEKkq8EolhaGLFE3XFs0oOvj0RCeQRdOU\n1dVmHXxyPXifiU4wnq5zMqlv0fjWcyy1rEJXKZqkwLcdwx5tB+9r0SgHn02TDr6OpQDjSX5U/fBU\niiYMeYOsStH0nLoWjW76kU9XFs1YO/i2LBp18PkM0aIZrcBnWTSKSfrTtINPClKemK2vbx0cH4so\n1RlklcA3Z4gxyVF48E0sGgl8NlWjfEnqWAowHlGqUtuNja1WllI09eniln3q4FsgvgdiEqVomjHP\nFM3QqdLBJ9ehAaVo6hLPsdiZUkTl4BeAvBx88iRybrugSODzaZqDlwefT5Xapm0FWTT1yLJnoB2L\nRjn4wORZNMnp4Jubs8k0MRL4fJrOZFWKJp8qFk16YFApmnpk2TOw+CmaUXTwWTn4tEWT7hZBAl9E\nkw4+eTIpB7+dOhZNjHLw9chK0IBy8AuBT4omnaAB5eCLCJmikUWzFVk03dOlRaNB1sD4pGiyOnjl\n4PPpwqIZq8CvrkYNR9YxmaaJRTOGWvqSZ9EseopmFBaNz0QnWTT+ODfrMuuIvFI0xcTHpY8Pnxam\nsjENpWiymZdFI4EPgI9FI4H3Z309iuUtLdWrjyyaYuL6+Ng0smjCMFSLplTgzexyM7vHzO4zs2sL\ntrvKzDbN7NKwu9icuhaNBD6buGvMmizmg1I0+TgXjfucd55fbZWiCUNXKZpe5eDNbA/wQeBy4AXA\na8zskoztzgHeCtzWxk42RRZNWGJRyZosVkY889LnIB9jBx8LwFlnqYPvki4tmi5z8GUd/IuBo865\nB51z68CNwBUZ2/0a8B+AVaB399vxmeiU7hZBAp9HLCq+NkKSqjMvxybwsdD41raKBy+Bz2esFs1B\n4IHE42PT555gaslc6Jz77PSpmtmK9vD14LNikhL47TSxaOpaCjAOUapaW996xnfCiifyjaGWVSiy\naBZZ4MtikoVibWY7gPcC/zz5dNOdCk3eRCcfD/7kyfb3b9GIux2z6h18ulPSRKetVL068q1n+gp1\nTPe49SHPomk7JjnviU7HgEOJx4fY2tGfA1wErFh0lPxt4M/N7Ern3J3pFzt8+PAT3y8vL7O8vFxr\np6tSN0WjHHw28clQR+CrWgpji0lWHd/IqmfW76VruWNH9P5tbg7/Fog+9MmiWVlZYWVlJcjfKxP4\n24GLzexC4GHgauCN8Q+dcyeAJ8WPzewvgLdliTtsFfguUUwyLEkfvU4HL4smn2QHH9KiyTq+43pK\n4Lu1aPbvj77Pe6/Sze+RI0dq/71CD945NwGuAW4B7gZucs7daWZHzOzK2n+1Y/JSNIpJ1qOqCGX9\nboxikltJevAhUzR5Aj/0evpS1MEv8mJjpUsVOOduBm5OPXddzrYvC7RfQfFdi0YC70eXFs0YO/iq\nFs2+fbPHdTp40V1Mslc5+KGgFE1YqnaZSWTRFDMPi0YMN0UzGoFP5+DjtEG8loosGn+6smjGKPBt\nWTRZV6hjqKcveRZNGymaPk10GgRZHfyOHVGBY4GSwPvTZCarUjTFhEjR5HXw6SvUMdTTFy02tsBk\n5eBha5ckgfcnpEWjHPxWmlo0efWURVPMvCyaeefgB0FWBw9bT6I8gdcNP7aTnOikFE1YlKKZD6NN\n0QyBPIFPXgZneZSa6JRNyBTN0lI02Sa+J26SsXbwsUVz+nT59lUsGnXw+ciiWWCKOvikRaMUjR9N\nLZqkIJn5i9IYBKmORSOBb45SNAuKc9EbVNeikcBvp2mKJt0pSeBn1LFoFJNsTlcpGuXgAxN35lkL\nKiUtGgm8PyFTNOCf/BhD6iMWmtApmqybyo+hnr7IollQsjLwMUrR1COkRQPq4JMk14OXRdMdXVo0\nysEHJM9/B1k0denKolGKpnz7phbN0Ovpi1I0C0qRwMuiqUfIFA3kdzJj7eCbWDS+tYRx1NOXoVo0\ngxf4vElOsLVLyltsTDn47TRdLjjLovGZnDMGQQph0WiiU3U00WlBKbNoimKSysFn0/SOTkrR5FPn\njk5K0TRnHjf8iO2fNu+qNWoPPrkmvCwaf5IpmqoevG+KZnMzOvCTk5/GkPqI66MUTbfM45Z9ZpHm\ntKkx6uDlwVcm5D1ZIbvOY+04kxZNWW2diy7x1cE3p6sOPm0Zx+9XXtKvKRJ4CXxlQt6yD7LrnLe8\n7dBTH1USSqurUe2Sl/dK0VRnczOqT5bItpmigfY1ZtQCL4umHskUTdPFxkAdfJIqFk2TOQUwjnr6\nsLoaiXuWD96mBw8S+Mb4TnTSLfv8aWLR+HrwYxWkKhZN1VnBY6ynD3n2DMw8+FADoWk9ajvIMepB\n1rIUzdJSJPybm+3u46IRcj14yD7IxypIVSwa31rCeOvpQ15EEqJB/h07wmlA1x384AW+KAdfZtEU\nrXQ4ZpqkaHxz8GMVpKYWTV5Tolv25ZOXoIkJmaTJEvg2s/CDF/gmg6ygLHwWVbPaSapYNGOM9SVn\nsp45U9w5ZtUyrykZaz19KLJoINwHoXPbP2jVwTekqcCrg99OaItGKZoZyQHsWOTLtk1TZUxj6PX0\nociigXB1WluLXssn9RSKUQt8mUUDEvg0Gxuz9fV37pytt++LUjTFJMWm7AM0r/NUPatRZtGEqlOW\nFkngG+LbwWd1jCCBT7O6Ousw4y6zig+vFE0+zm0VmzIfPq/zVD2r0ZVFkzUeKIFvSJMUDUjg06RF\npYpN49wsc5xEghSxvj6bvg7lSZoQFs2Q6+mLj0WjDr6nFOXgZdFUJ93tVFkTPhb39M21JUgRVT88\nZdGEoayDDzUYnaVFysE3RIOsYUl3jVVu25fXKeXl4MeW+kgLTV2LJqueWmwsH3nwC0wIgdea8DOa\nWDRVLIUxpmjS9enCohlyPWNuuQWOH8//eZcpGgl8YHxv+JEn8Pv3w+OPt7d/i0YTi+bUqaieaTTR\nKaLqh6fq6cd118Gtt+b/vKtB1jyB10SnBvjGJPNSNOedBydOtLd/i0YTi+bECThwYPvz8owjqlo0\nqqcfJ04Un8OyaBaYphbNeecVX96NjSYWzfHjUT3TSJAiqlo0qqcfx483t2gk8D2laUzywAF18Emy\nLBp18GGoWlvV0w+fDr6LFI1y8C3QdCarLJqtZFk0vh78iRPVOs6xpT7SnaSPReNbz7GmaM6cgR/+\nUBbNYCkT+MlkNt1eFk05XVk0StF0Y9EMuZ4wE/a+pGiUgw9M0USnnTujLmZtLV/gZdFspQ2LRuvB\nR4SyaFTPGfG528SiGXwHb2aXm9k9ZnafmV2b8fO3m9lRM7vXzD5vZs8Mv6v1KOrgYWYxKEXjR5cW\nzdgEqU2LZoz1hFnnLosmBzPbA3wQuBx4AfAaM7sktdltwKXOuYuBjwLvDb2jdSkT+LhLkkXjR1OL\nRoOC+VSdY6AUTTknTkT1UIomnxcDR51zDzrn1oEbgSuSGzjnvuCciw/FLwEXht3N+hRNdIKtAq8U\nTTlNLZo8QdLEnGwPPq+2zqmePpw4AU97Wr8tmnlPdDoIPJB4fGz6XB5vBD7ZZKdC4mvRKEXjh1I0\n7VHFoplMokXbfBcbG2uK5vhxP4Hv4pZ98+jgMyRtG873xczstcClwGVZPz98+PAT3y8vL7O8vOz7\n0rWRRROWLIvm9Gm/35VFU0wViybPngHVM8mJE/D0p8Mdd+Rv05VF45uDX1lZYWVlpfkfxE/gjwGH\nEo8PsbWjB8DMXgG8A/hp51zmZ1JS4LuiqcDLotnKZLJVpPfuhUcf9fvdqrntscX6qlg0eQkaUEwy\nyYkTcPBg1IRkXcWAn0XT5WJj6eb3yJEjtf+mj0VzO3CxmV1oZruAq4GbkxtMB11/D7jSOfdI7b1p\ngaYWzdlnzw4O0cyiaToouLQUvQ/O+5pysahi0aiD9+P4cTj//Og8Pnkye5tRp2iccxPgGuAW4G7g\nJufcnWZ2xMxeOd3sN4D9wMfN7C4z+7PW9rgiRTl4mHVJeTHJHTvg3HPhscfa28dFom6KpmxQ0EeQ\nzGYiP0SqWDR5tQQJfJL4SufAgXyrtcsUTdcTnXwsGpxzN5Pq2p1z1yW+/9nA+xWMphYNzAZazz+/\nnX1cJOqmaIoGBX0n5sDsZMt7rxaZUBaNJjrNiD8Ii8IS807RzDsmudBUsWiy/DnQQGuSuhZN3gAr\n+KdoYNjJj7YtmrGmaA4cKBf4oaZoBi/wVXLweV2hBlpn1LVoyiwFn9w2DLvrrGrRFH1gpuuZN2g9\n1FrGxMddXyyaPubgF5qQFo2ob9FU9YzzxkSGnPyoatGE8OCHWsuYUBaNbtnXU3wsGh+Bl0UTUfeO\nTnUsmrF38ErRNCdp0eSdw12laLQefAv4dPBFi42BLJokde/JGqLjhGGLUhX7K1QOfqi1hK3Jrbxz\n2Lmo7qONSS46smjCUteDD9FxwrBFqcp68Orgy3n88ej/uHt3/jm8tjZbNjwPCXyPKcvBx5fBGxtR\njC8LWTQzmnjwsmiKqVJb5eDLSR5zeedwmT0DYVM0uuFHYHw6+NOnozfRLHsbWTQz6sYkiwQpL7et\nmGS4HPwYFxtLHnN553DZACuog+81PgJ/6lTxxBl18DOaWDRVOvixpmi6XmxsqLWErcdcnkVTFpEE\npWh6jW8HXybw6uAjuopJjtWiyUooZa29I4umnHQHn2fRqINfYMomOu3ZUy7wsmhmtGHRxOvLbG7O\nnhujwKe7yaWl2T2Dk2xuRgtnnXtu9uvohh8RyWMur0nz8eA10anHyKIJh3PRwZgW+DNntopzFkUW\njdn2TmaMAp/VTWbZNCdPwr59+cesOviIkBaNcvA9xLnojfER+KKYlDr4iPgATaaNzPy6+KIOHiTw\nkN1NZllgVWsJ4xR4WTQDF/g4iZGXjgE/i0YefETe5WxbAj+mFM3GRvSVFoCsJE1RggZ0y76Y5DF3\n1lnR/zVth3Qdk5TAB6TMngE/i2bv3ugE8b2xxVDJu5z1GWgtsmhg+4E+thRNXNt0M5Jl0RQlaEAp\nmpjkMWeW3agpRbPAlE1yAr8UjZlsGsi/nC0T+LJBQdie3R6bRZPXSeZZNEUflulaxlZluoMfai1j\n0leNWTZN1xaNJjoFxKeD97FoQDYN1LdoTp6E/fuLxznG7sHndZJZFk3VDn5zMxo3Sc/UHmotY9JX\njVnncB9SNBL4moSyaEBJGqhv0ZT57yCBL7o6Sn94hhqwHvo9btN1amLRSOB7SFkGHmZvbpnAy6Kp\nb9H4CnxyAGyMAh/KovGt5dDvcdtHiyatR/Frt/UhO2iB97VooNg+AHXwUN+iKRtgBaVo2rRoim5H\nOdR6QjiLJlSNshpOs0jk2+riRy/wvh28PPhuLZqxpWjatGiK7nUw1HpCWIumrRQNtGvTSOBl0XjT\ntkUjD37783UtGp9awnDrubkZja2dc87suT5aNCCBr00Vi0aDrOUUWTRFAl/XohmbwPvWto5FMzaB\nf+wxOPvsrdZUH1M0IIGvjU8OfteumQ9WhCyaYoumyIP36eDHnoOvUtuyevrWEoZbz6wa9SFFk6VH\nbWbhBy997XqlAAAJzElEQVTwZR28WfQGy6Ipp4lFow6+GFk0YcmqkSyageEj8BBdoilFU04Ti6aO\nBz+mFE1Ii2ZpKfKg4+jdGFM0WTWad4pGAh8YX4FXB+9HmxbN2FM0IS2adPRujCmarBplncNdpWic\ny38fJPA18ZnoBH4CLw++vkXjO8g69olOPrVdW4tE6eyzi18vWc8xWjRZx1zWVXhXFs3aWvQ6WSvb\ntnnTj0ELfBWLRimacupaNIpJluNb2xMnokXbipbAhq31HKPA+w6ydpWiKWo21cHXRBZNWLq0aMYm\n8L619RmwBgl8kcAnlwXoKkVTpEUS+JqEFPhzz91+cIyNti2aMQu8b219BqxBAp91zO3eHdXl8cdn\nz3Vp0UjgA+OTgwe/FM2uXbOlhcdKmxbN2HPwVSwaH4FP1nOMAp9Xp7RN05VFU6RFysHXJGQHD7Jp\n6lg0a2uR/7h/f/FrZ6Vo8mKSQ0x9tGnR5NUShlvPvDqlz2EfiyZEjdTBt0BogR/7QGsdi6bOoCCM\ns4OXRROOvDqlz2FZNAtMGwI/5g6+rsBXFSQYp8D7rAdfp55jFPgqFs2oBd7MLjeze8zsPjO7NuPn\ne8zsxuk2XzKzp7ezq9UJGZMEWTSrq9XXg69jKeTdQxSGK0hF68ErRVOdIosm2cHnHdNJBivwZrYH\n+CBwOfAC4DVmdklqszcD33PO/Tjwm8D729jROoSc6AThLJqVlZXmLzIH6nTwZZZCXIvkZI+8e4jC\ncAVpMoG/+quVbc83sWgWeaJT03OkyKKZRwdfloOf10SnFwNHnXMPOufWgRuBK1Lb/ALwken3fw78\nlFmZ49oNfbVoxiTwZZZCUuDH3HFOJnD//Svbnh+rRdP0HJFFE1Em8AeBBxKPj02fy9zGObcJPAo8\nOdQONiHkYmOQvRrdmKgTk/TJwMN2QRrb4liTSbYIp2urepZz5kz0f9+3b/vPkudwfC/UsuYuRI3m\nJfBlfWvQaT1XXhny1cq5/3544xvLt9u3L9sOSHP++XDDDfClLzXbr69/Hb7ylWavMQ++8Q0466zt\nz+/bB9/8Zvb7++1vw8tfXv7ae/bAJz8Jf/3X0cmUdzLs3g2f+1z3x1Lb3HsvPPe525/ftw+++93Z\n//e22+BNbyp/vT174B3vgN/5HXjwQXj+87O3270b3vWu6LjuE03OkbW1SMizfITzz4cPfQjuuSeK\nPmYdz2niDr7JMffII/lR4T174Prr4dOfjh4/5znwW79V/28lMVcwNdPM/gFwrXPuldPHbwd2O+fe\nldjmf063ucPMdgAPAU+ZdvPJ1xrxHFAhhKiPc66W7V3Wwd8OXGxmFwIPA1cD6Z74s8A/A+4AXg3c\nmhb3JjsohBCiHoUC75ybmNk1wC1Efv1HnHN3mtkR4A7n3KeA64GPmNk9wEngl9reaSGEEOUUWjRC\nCCEWl9ZnspZNlBoyZnbIzD4//f9/3cz+1fT5HzGz/25mXzWzW8zMIxcxDMxsyczuMrNPTR8/08xu\nndboj83MI/e0+JjZATP7r2Z2t5ndb2YvGetxYWZHzOz/mNnXzOzjZrZvLMeFmd1gZg9NHZD4udzj\nwMzeb2ZHzezOjDlJ22hV4D0nSg2ZM8CbppPAfgJ4g5m9EDgCfMY59wLg5unjsfBW4D5mCa33A++Z\n1uj/EU2cGwO/D9zknHshcBFRTUZ3XJjZs4DXARc7554LbAC/yHiOiz8g0sckmceBmV0FPM05dxHw\ny9PfLaTtDt5notRgcc495Jy7d/r9KeCrwIVsnRz2R4ykJmZ2kOj//qHooS0BL3HO/dl0k1HUwsz+\nFvAi59zHIJo/4px7jHEeFz8A1oD9ZrYT2Af8X0ZyXDjnvgD8TerpvOPgivh559xdwM7pOZVL2wLv\nM1FqFJjZM4C/B3wReJJz7lEA59wj9GRiWAe8D3g7EKesngw8kvj5g4zj+Pgx4Ptm9idmdq+Z/Wcz\nO4cRHhfOuR8Av00k6t8FjgP3Ms7jIibvOLiQinratsBrBBcws7OBjwNvnXZqo8PMXgk8PO084sjs\nWKOzO4g+7H/TOXcxURf77+e7S/PBzH4U+BXgGcDfAc4Gfnae+9Rz0udMoca2LfDHgEOJx4fY+gk0\neKaDQ58A/kvikvP7ZnbB9OdPIppjMHR+CniVmX0L+BjwM8B7gAsS2xwkOmaGzgPAg86526ePPw68\nCHh4hMfF3we+7Jx7dGrj3gT8NOM8LmLy9CGtp6V1aVvgn5goNRW6q4kGDUbBdNG1DwP3Oefel/hR\nPDmM6b+f7XrfusY592+dc4ecc88E/inwv5xzrwNuM7N/ON1sLLV4AHjEzJ49feoVwP1E58aojgvg\nG8BLzOys6fnyCuBrjPC4SJCnD58FXgtgZpcCG865B4teqPUcvJn9PNEywvFEqXe3+gd7hJm9FPg8\n0eBqXOh/A/xvogHnpxAlBK52zo1mGTMzuwx4m3PuVWb2TOCjRJfmR4HXOedaWnqpP0zTVB8iGlT8\nDtGJa4zwuDCzw0T//03gLuBfAE9lBMeFmX0MuIzoiuUh4J3AJ8k5DszseuBlwCrwBufcnYWvr4lO\nQggxTAZ9yz4hhBgzEnghhBgoEnghhBgoEnghhBgoEnghhBgoEnghhBgoEnixMJjZM5LLqqZ+dsTM\nfmb6/a+YWebdNs3s983seRX/7qnqeyvE/FEOXiwM0wXbPjVdQrZou28BfzdesCnA3z3pnDsnxGsJ\n0SXq4MWisWRmvzddhXHFzPYDmNkfmtlVZvYWokWr/sKiG8JvYfo7l06/P2Vmvz69AcldZvbU6fPP\nnj7+ipn9eur33zm9EcP9Zvbu6XP/yMz+x/T7p1p0c5fBrwQp+o8EXiwaPwZcP12F8SHgH0+fd4Bz\nzn2AaNnZZefcyzN+P3nJug/4onPuEuBzzG4o/wHgN5xzP0G0jC0AZvYq4MLpjRguAn7czF7hnPtT\n4Htm9mbgPwHvdM6NYaEw0XMk8GLR+FZ8ExXgK2xdXa8qZ5xz/y3jtX4S+JPp93+c2P7ngJ8zs7um\n2z+baJlbgLcQrTM0cc7d2GCfhAjGznnvgBAVWU18v0GzJiW5eNVm6rXyBqd+zTl3Q8bzh6b78xQz\nM6fBLdED1MGLIfJDYH+D3/8y0dLWAP8k8fwtwOvNbC+AmT3FzC6Y3mruw0TLIH8N+NUGf1uIYKiD\nF4tGujPO6pQ/TDTI+p0cHz7rd13i8VuAG83s7cBn4uedc58ys+cDd5rZGWAC/BJwDfB559yXzeyr\nwO1m9mnn3Ner/ueECIlikkIIMVBk0QghxECRwAshxECRwAshxECRwAshxECRwAshxECRwAshxECR\nwAshxECRwAshxED5/0nc6aTzDuPWAAAAAElFTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f93105d1c90>"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.hist([val for val in trainer.network.predict(train[0])[4] if val > 0.1 ], bins = 50)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 12,
       "text": [
        "(array([ 1.,  0.,  1.,  0.,  0.,  0.,  0.,  1.,  0.,  0.,  1.,  0.,  0.,\n",
        "         0.,  1.,  0.,  0.,  0.,  0.,  0.,  0.,  1.,  0.,  0.,  1.,  0.,\n",
        "         1.,  0.,  1.,  0.,  0.,  0.,  1.,  0.,  0.,  0.,  0.,  1.,  0.,\n",
        "         1.,  0.,  0.,  0.,  1.,  0.,  0.,  0.,  0.,  0.,  3.]),\n",
        " array([ 0.11243238,  0.13007094,  0.14770949,  0.16534804,  0.1829866 ,\n",
        "         0.20062515,  0.2182637 ,  0.23590226,  0.25354081,  0.27117936,\n",
        "         0.28881792,  0.30645647,  0.32409502,  0.34173358,  0.35937213,\n",
        "         0.37701068,  0.39464924,  0.41228779,  0.42992634,  0.4475649 ,\n",
        "         0.46520345,  0.482842  ,  0.50048056,  0.51811911,  0.53575766,\n",
        "         0.55339622,  0.57103477,  0.58867332,  0.60631188,  0.62395043,\n",
        "         0.64158898,  0.65922754,  0.67686609,  0.69450464,  0.7121432 ,\n",
        "         0.72978175,  0.7474203 ,  0.76505886,  0.78269741,  0.80033596,\n",
        "         0.81797452,  0.83561307,  0.85325162,  0.87089018,  0.88852873,\n",
        "         0.90616728,  0.92380584,  0.94144439,  0.95908294,  0.9767215 ,\n",
        "         0.99436005]),\n",
        " <a list of 50 Patch objects>)"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAAEACAYAAACnJV25AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxmKXpMZY7JLUGItdkhrz/1n1F+f8VmbNAAAAAElFTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f3de003e150>"
       ]
      }
     ],
     "prompt_number": 12
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "plt.plot(trainer.network.predict(valid[0])[163],'bo')\n",
      "plt.plot(ds_output[cut+163],'yo')\n",
      "plt.xlabel('hit index')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 25,
       "text": [
        "<matplotlib.text.Text at 0x7f40ff6dd4d0>"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEPCAYAAABIut/fAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAHnpJREFUeJzt3X2UXHWd5/H3l3SnO4QHkQcDSThi7yAMDBrGEdjRpCdC\nOjsNJDO4ONI6qyMHZNdORhnIGBLTHEDJcFbHbl0dfBgdiDEzCgnQkoS1aRqEaDCcYAJhjw2jSZMn\nwkPE0G0evvtH3e5UVVd1VXXfqrp97+d1DoeqW7++9a1f/eqbW7/7u98yd0dEROLnmGoHICIi5aEE\nLyISU0rwIiIxpQQvIhJTSvAiIjGlBC8iElMFE7yZfdfMdpvZr0Zo025mW81sk5nNCDdEEREZjWKO\n4P8VmJvvQTO7CjjT3c8DPhW0FxGRKiuY4N39ceC1EZr8JXBP0PYZoMbMpoUTnoiIjFYYc/DTgO1p\n93cE20REpIrCOslqWfdV/0BEpMpqQtjHDmA68PPg/rRgWwYzU9IXERkFd88+iC5KGAn+J8DHgB+Z\n2YXAYXfvy9UwCoXNFixo4q//ev2w7fff38RXv7q2IjG0tbXR1tZWkeeKuosuamD58heHba/k+xEV\nYYyLKIzvMEThMxKVvjQbVW4HilsmuRJ4Eni3mW03s78zs+vN7HoAd/8x0GdmW4FvA58cdTQVMH/+\nAlasaMjYdu+9Dcyb11qliJLtnHMu0vsRIo3v8MShLwsewbv7R4to85lwwim/2bObAbj//g6gH6jn\nmmtah7ZLZZ111tnMnNmi9yMkGt/hiUNfWqWmTczMozBFEwXd3d00NjZWO4xIUF8cpb44Sn1xlJmN\neg5eCV5EJMLGkuBVi0ZEJKbCWEUjEprOzh7a29czMFBDXd0hFiyYQ3PzzGqHJTIuKcFLZHR29rBw\n4Tp6e+8Y2tbbewuAkrzIKGiKRiKjvX19RnIH6O29g46OR6oUkcj4pgQvkTEwkPsLZX//hApHIhIP\nSvASGXV1h3Jur68/XOFIROIh8Qm+s7OHpqYlNDa20dS0hM7OnmqHlFgLFsyhoeGWjG0NDYtpbb2s\nShGJjG+JPsmqk3rRMtjnHR1L6e+fQH39YVpb5+q9EBmlRF/o1NS0hPXrb8+xfSlr195WhYhERDLp\nQqdR0kk9EYmzRCd4ndQTkThLdILXST0RibNEz8FD6kRrR8cjaSf1LtNJPRGJDFWTFBGJKZ1kFRGR\nYZTgRURiSgleRCSmlOBFRGIq0aUKRERGMt5/gEYJXkQkhzjUqtIUjYhIDnH4ARoleBGRHOJQq0oJ\nXkQkhzjUqlKCFxHJIQ61qlSqQEQkjyjUqlItGhGRmFItGhERGUYJXkQkppTgRURiSgleRCSmlOBF\nRGJKCV5EJKZUbEwkZsZ7BUQJjxK8SIzEoQKihKfgFI2ZzTWzX5nZc2a2KMfj55jZz81sS9BmXnlC\nFZFC4lABUcIzYoI3szrgG8Bc4ALgw2Y2I6vZEuC77n4+cBXwtXIEKiKFxaECooSn0BH8RcBWd+9z\n90PAKqA5q8124MTg9tuA34QboogUKw4VECU8hRL8NFIJfNCOYFu6LwH/w8y2A51Aa3jhiUgp4lAB\nUcJT6CRrMdXBvgx8292/YmYXA/cC5+Vq2NbWNnS7sbGRxsbG4qIUkaIMnkjt6FiaVgFxrk6wjiPd\n3d10d3eHsq8Rq0ma2QeBRe5+eXD/JmCiu9+R1mYb8CF37wvu9wKXuPuerH2pmqSISInKWU1yI3C+\nmU01s1rgauDhrDa9wKVBIOcCk4F9owlGRETCM2KCd/d+4AZgHbAZuM/dN5nZrWZ2RdDsc8CnzWwr\n8GPgWnfXGR0RkSrTD36IiESYfvBDRESGUYIXEYkpJXgRkZhKfLGxrq5OVq9ux2wA9zrmz1/A7NnZ\nF+smT7UqEur9CJf6s3Ki2NeJTvBdXZ2sXLmQlpbeoW0rVqRuV/uNqaZqVSTU+xEu9WflRLWvEz1F\ns3p1e8YbAtDS0suaNR1ViigaqlWRUO9HuNSflRPVvk50gjcbyPNIf0XjiJpqVSTU+xEu9WflRLWv\nE53g3evyPFJf0TiiploVCfV+hEv9WTlR7etEJ/j58xewYkVDxrZ7721g3rxkF8SsVkVCvR/hUn9W\nTlT7OvFXsnZ1dQbzZP1APfPmteoEFKkTrR0dj6RVJLysYqto9H6ER/1ZOeXq67FcyZr4BC8iEmUq\nVSAiIsMowYuIxFQiL3Sq1lWaIiKVlLgEX62rNEVEKi1xUzTVukpTRKTSEpfgq3WVpohIpSUuwVfr\nKk0RkUpLXIKv1lWaIiKVlsgLnap1laaISKl0JauISEzpSlYRERlGCV5EJKaU4EVEYkoJXkQkphJX\nqkAkSVR3KdmU4EViSnWXRFM0IjGlukuiI3ipOk0jlIfqLokSvFSVphHKR3WXRFM0UlWaRigf1V0S\nHcFLVWkaoXwGvwF1dCxNq7s0V9+MEkQJXqpK0wjl1dw8Uwk9wTRFI1WlaQSR8lE1Sak6lW8Wya+s\n5YLNbC5wFzAB+L67L8/R5mrg80GbLe5+TY42SvAiIiUqW4I3szpgG/ABYDfwFHCduz+T1uY9wN3A\nbHf/vZm93d1fzbEvJXgRkRKVsx78RcBWd+9z90PAKqA5q80nga+5++8BciV3ERGpvEIJfhqwPe3+\njmBbuncD7zWzp83sl2Z2ZZgBiojI6BRaJlnMnMoxwDtJHe1PB540syd0JC9JoDILEmWFEvwOUkl7\n0HQyj+gJ7j/h7oeB/zSz54CzgQ3ZO2traxu63djYSGNjY+kRi0SEyixIOXR3d9Pd3R3KvgqdZK0n\ndZL1z4E9wJPA9e6+Ka3NXwHz3P0TZnYKsBl4r7vvzdqXTrJKrDQ1LWH9+ttzbF/K2rW3VSEiiaOy\nnWR1937gBmAdqcR9n7tvMrNbzeyKoM39wD4z2wo8AfxjdnIXiSOVWZCoK1iqwN0fBh7O2rYs6/6N\nwI3hhiYSbSqzIFGnUgUio6QyCxJ1KlUgMgYqsyDlVtZSBWFRghcRKV05r2QVEZFxSgleRCSmlOBF\nRGJKCV5EJKaU4EVEYkoJXkQkpvSj2yISGarOGS4leBGJBFXnDJ+maEQkEtrb12ckd4De3jvo6Hik\nShGNf0rwIhIJqs4ZPiV4EYkEVecMnxK8iESCqnOGLxEnWbu6Olm9uh2zAdzrmD9/AbNnN1c7rHFL\n/XmU+iI8zc0z2bLlKR54oIHa2sMcPDiBK6+8TidYxyD2Cb6rq5OVKxfS0tI7tG3FitRtfRBLp/48\nSn0Rrq6uTn79629xxx0vDm1bseJbdHWdr/4cpdhP0axe3Z7xAQRoaellzZqOKkU0vqk/j1JfhEv9\nGb7YJ3izgTyP9Fc0jrhQfx6lvgiX+jN8sU/w7nV5HqmvaBxxof48Sn0RLvVn+GKf4OfPX8CKFQ0Z\n2+69t4F581qrFNH4pv48Sn0RLvVn+BLxk31dXZ3BPF4/UM+8ea06aTMG6s+j1BfhUn8Op99kFRGJ\nKf0mq4iIDBP7dfClUKlSEYkTJfiASpWKSNxoiiagUqUiEjdK8AGVKhWRuFGCD6hUqYjEjRJ8QKVK\nRSRutA4+TWdnDx0dj9DfP4H6+sO0tl6mE6wiUlW60ElEJKZ0oZOIiAyjBC8iElNK8CIiMZWIK1lV\ngkBEkqhggjezucBdwATg++6+PE+7q4D/AN7n7ptCjXIMVIJARJJqxCkaM6sDvgHMBS4APmxmM3K0\nOx5YCGwoR5BjoRIEIpJUhebgLwK2unufux8CVgG5qu/fBtwJDACjWs5TLipBICJJVSjBTwO2p93f\nEWwbYmYXAlPd/SfBpkgtdlcJAhFJqkIJfsRkbWbHAF8G/iF981iDCpNKEIhIUhU6yboDmJ52fzqZ\nR/THA+cB3WYGMAV4wMyuyHWita2tbeh2Y2MjjY2Nowq6FIMnUjs6lqaVIJirE6wiEknd3d10d3eH\nsq8RSxWYWT2wDfhzYA/wJHB9vlUyZvYocGOux1WqQESkdGUrVeDu/cANwDpgM3Cfu28ys1vN7IrR\nPKGIiFSGio2JiISgXBdUjuUIPhFXsoqIlFNUL6hULRoRkTGK6gWVSvAiImMU1QsqNUUjJVPxNpFM\nUb2gUgleShLVuUaJn/F0ILFgwRx6e2/J+FykLqicW8WolOClRPnnGpdG9sMn4894O5CI6gWVSvBS\nkqjONUq8jMcDiebmmZGLTSdZpSRRnWuUeNGBRDiU4KUkKt4mlaADiXBoikZKEtW5RomXqJ60HG9U\nqkBEIqmzs4eOjkfSDiQuS+SBxFhKFSjBi1TAeFryJ9GiWjQiETbelvxJfOgkq0iZRbVOicSfErxI\nmWnJn1SLErxImWnJn1SLErxImenaAakWraIRqQAt+ZPR0jJJEZGYKtuPbouIyPilBC8iElNK8CIi\nMaUELyISU0rwIiIxpQQvIhJTSvAiIjGlapJpuro6Wb26HbMB3OuYP38Bs2c3VzusqlBfhEv9WTr1\n2dgpwQe6ujpZuXIhLS29Q9tWrEjdTtqgUl+ES/1ZOvVZODRFE1i9uj1jMAG0tPSyZk1HlSKqHvVF\nuNSfpVOfhUMJPmA2kOeR/orGEQXqi3CpP0unPguHEnzAvS7PI/UVjSMK1BfhUn+WTn0WDiX4wPz5\nC1ixoiFj2733NjBvXmuVIqoe9UW41J+lU5+FQ9Uk03R1dQZzfP1APfPmtSb2hI76Ilzqz9Kpz1JU\nLlhEJKZULlhERIZRghcRiamiEryZzTWzX5nZc2a2KMfjN5nZVjPbYmY9ZnZW+KGKiEgpCiZ4M6sD\nvgHMBS4APmxmM7KabQAudPfzgR8AXw47UBERKU0xR/AXAVvdvc/dDwGrgIxT2e7+uLsPXpnwM2Bq\nuGGKiEipiknw04Dtafd3BNvyuR5YM5agRERk7IopNlb02kYzawEuBGblerytrW3odmNjI42NjcXu\nWkQkEbq7u+nu7g5lXwXXwZvZB4FF7n55cP8mYKK735HV7lKgHZjp7q/k2I/WwYuIlKjc6+A3Aueb\n2VQzqwWuBh7OCmAG8E3gilzJXUREKq9ggnf3fuAGYB2wGbjP3TeZ2a1mdnnQ7J+AycCPzOwZM1td\ntohFRKQoKlUgIhJhKlUgIiLDKMGLiMSUEryISEwpwYuIxJQSvIhITCnBi4jElBK8iEhMKcGLiMSU\nEryISEwpwYuIxJQSvIhITCnBi4jElBK8iEhMKcGLiMSUEryISEwpwYuIxFQxP7otMdbZ2UN7+3oG\nBmqoqzvEggVzaG6eWe2wRCQESvAJ1tnZw8KF6+jtPfr76b29twAoyYvEgKZoEqy9fX1Gcgfo7b2D\njo5HqhSRiIRJR/AJNjCQ++3v759Q4UikGjQ9F39K8AlWV3co5/b6+sMVjkQqTdNzyaApmoTp7Oyh\nqWkJjY1t7N27iylTPpfxeEPDYlpbL6tSdFIpmp5LBh3BJ0iuo7YpUz7FhRf+L44//lTq6w/T2jpX\nR3AJoOm5ZFCCT5BcR227dn2H97xnKWvXtlUnKKkKTc9VTva5jksuOYOnnnq5Iuc+lOATREdtMmjB\ngjn09t6S8Q9+anpubhWjip/h35p76Or6AYcOfXOoTTnPfSjBJ4iO2mTQYDLp6FhKf/8ETc+VyfBv\nzeszkjsMnvtYqgQvY6OjNknX3DxTCb3Mhn9rruy3aCX4BKnmUVsYa661bluqrdQxOPxbc4W/Rbt7\nRf5LPZUk0UMPPeYNDYsdfOi/hobF/tBDj1V0HyJjMZoxOPxvHvOamuuz9vH5EfcR5M5R5V1L/X35\nmZlX6rkkWpqalrB+/e05ti9l7drbKraPqEraN5Px+npHOwY7O3vo6Hhk6FvzxRefzoYNO9O+RV82\n4us3M9zdRhOzpmhGMF4HYtSEsXonriuAknZF6Xh+vaMdg9U816EEn8d4HohRk3v1Tg9btjxPY2Pb\nKOcyU8b7CqD8V5SWZ1VFtY3n11vKGIzKwaESfB7jeSBGzfDVOz3U1PyAfftW8dhjqS2F/vGM0gqg\nMD+8oz0qTI9h//4dwEROOOG0yH/THP56e4D1bNiwg6amJZGOPdc4njTp6/T1nZ4Re5QODpXg84jr\nlEDYikl22at3tmx5nn37VmW0KfSPZ1TWbYf94R3NN5PMGHqAdUD1k0kxMl/v0djfeAPWr4927Olj\ncMeOPbz4ovHWW6vYsgW2bDkae5QODmOV4MM8shrpgxeVr1+VlOs1AyMmu3z91NjYNnTkni7Kc5mD\nRvrwDj5eyrgYzTeTzBjWk57c0+PJfu5qjtvB5+7r28ukSZ/mrbe+WVLs+fZXymsJ4/UPjsGmpiVs\n3Zp5wnUw9lIODsv9nhRM8GY2F7gLmAB8392XZz1eB/wb8MfAfuAad/9NaBFSXCeEfWSV74N38cXT\nIvP1q1Ly9e0JJ7xGb+//yWibnuzy9VNU5tNH8+HK9+HdsWNP0eMi+3k/9rGpbNiQ+mYyON1y111d\ntLevzxlTZgzFJZNCn4+R+qLYfso3bbR//w527jyBXbu+HLTsYdKkj2BWy4EDhWPPte/s/T3++Ndp\naLiPM844bsz5oZjXO1ISL3Z8V2QqZ6Q1lEAd8BIwldRI2gjMyGpzI/DPwe35wJo8+ypx1Wn+daST\nJl3t55+/0OfMuWVo/eicObdkrC2Fxxxu8ZNO+tuMdqU+d1PTEp81a5k3NS3xhx56LMfzpP5ralqS\n8+/nzLnFZ81alhHDo48+Oqq+qJZ8r/mkk/425/bB15uvnzLf00eLWgsctlxrmqdM+TufMeMGnzVr\nmc+Y8amct08++eqcryv39sf85JOvznj/R1pL/cUv/vOI66wHx9NJJ30krU1x47H49yOzL8477zqf\nNCl73fbwtd+Z+3jMYXHBGPP1ZVPTkozPyPD40veX/VzF9Nnw/LBs2dd9zpxbin69pfZnrvFdbM6i\nXOvgzWwmcLO7Xx7c/weg3t1vT2vz06DNL83sGGA3cJpn7djM/IIL3k5t7TEcd9xkDh48hlNPPZe9\ne5+ntvYIr732u6HH0m/v29fP9u2fZsqUp6iv382BAzupqZnKxIkncODAy0ycuJcTT5zMG28cYceO\n63K26++v4+DBE2lv/0xJ/zJ2dXWyenU7ZgO88MLLvP76fg4dgv37z6CmhqF979p1OtOmreWUU+qH\nYq+rq2Hv3oPU1Jw51O7NN3dz+umvsnfvfk48cULO1ztSvxTbZyO127//LaZOPZF3vesM3Ot417su\n4cUXn8p4jfX19Rw8eAxXXnk9ixYtorGxjV/84s8488x26usH6O+v47e/XUB9/T309388Y/uuXacz\nffpa6uoy++nAgf3U1MBpp73KBRe8m0OHTmfz5sfp69vH1Kknc+WV17Fo0aJh78Hy5ct54IF/obb2\nSEbs2bGW2mdg7N8/Le19PMCuXROZMmVy1vg5kjWWDrBnz2FOO+2Uodd88OAJHHvsVHp7L0vri/T9\nDQyNwRNPPIVt25qz+uwSzjvvHvr6XmHixIa0PnuZmpr9HHusY1bLzp3TOO64EwvEl4qnvb01Y6zn\neg937bqEc8+9h/r6Q/T1nZynL7Zx4MDJOWOqq5s81LcDA2+m9ee+jL/p79/Orl0fDz6bA0Nj4aST\ndvLGG0c4cmR47Bs3dtHW1gak1p8//vglabGn7y9/fJl99gIHDrw9aDfSe7otK/bUWJ85cwNr1942\nlBNee203zzyzM2fszc0zg3F7N7W1h9m/v5+pU08Y9pl79tkXePnlfDEdzVmXXz4LH+U6+EIJ/hrg\ng+5+Q3D/b4BGd/90WpsXgjZ7gvvbgFnuvjtrX97SAtdem7q/eTOsW2fcfLOzeTNs3Jh6LP320XY1\n3HzzoVG1G3TrrQ0cPPg+nnjih0V1TFdXJytXLqSlpbeI5y32daTa3XknnHJKafsb7XMVatfVVcNn\nP5u7z774xbfxoQ/9Iw8+uIna2l+ybFlvRn/u3Hk8Z575nyxe/Pqon/d734NPfAJWrGjgox/9KrNn\nNw89x/Lly/npT+9k8eLXR/V6yzfOJnDzzUe/bt999xm88MK7OOGEnUN9lL6/9D7btu1YzjnnQM52\nhcdF4fgG47n22rsz+vIDH/hIxns40ueltL4Ir2/TY+/p2TiU4N///muYPPkXBWMvZ+74/e/fz513\ntuTMCdmxA3nb5fvMjZSzfvazVaNO8IV+0SnUS0/Tg3/6abj5Zh+6PfhY+u2j7Q6Nut2gZct6OXLk\n6aJjXb26nZaW3iKft9jXkWr38sul72+0z1Wo3Wc/m7/PFi9+nQceuJvp01/KSO6Q6s8pU14cSu6j\nfd5BLS29rFnTkbHtgQf+ZWj/o3m9pfdZseMscy71uute5vjjn83oo/T9DVq2rJepU1/I267wuCgc\n32A82X2Z/R6O9HkprS/C69t8sdfWbigq9nLmjtran+fNCdmxj9Qu32cujJyVS6Ej+A8Ci9KmaG4C\nJrr7HWltfhq0eTptiuYd7n4ka1+h/mMhIpIUoz2CL7SKZiNwvplNBfYAVwPXZ7X5CfAx4GlgHvBU\ndnIfS4AiIjI6IyZ4d+83sxtIXY1wDHCPu28ys1uBp939QeBrwD1m9ivgd8A15Q5aREQKq1g1SRER\nqaxCJ1nHzMzmmtmvzOw5Mxu+Di7GzGy6mfUEr/8FM7s52P52M3vEzJ41s3Vm9rZqx1opZjbBzJ4x\nsweD+2eZ2VNBH/3QzGqrHWMlmNnbzOw/zGyzmT1vZhcndVyY2a1m9v/MbJuZ/cjMjk3KuDCz75rZ\n7mAGZHBb3nFgZu1mttXMNpnZjEL7L2uCD65y/QYwF7gA+HAxQcXIH4D/6e5/AvwpcK2ZvQe4Feh0\n9wuAh4P7SbEQeI6jK7TageVBH+0CPlOtwCrsW8B97v4e4DxSfZK4cWFm/wX4OHC+u58DHAY+SnLG\nxb+Syo/pco4DM7sKONPdzwM+FfztiMp9BH8RsNXd+9z9ELAKaC7wN7Hh7rvdfUtw+03gWVJXBf8l\ncE/Q7F4S0idmNo3Ua/926q5NAC5299VBk0T0hZmdDLzX3VcCuPsRd99PMsfFq8BBYLKZ1QDHAr8l\nIePC3R8HXsvanG8cNA9ud/dngJrgM5VXuRP8NGB72v0dwbbEMbN3An8GPAGc6u77ANz9FeC06kVW\nUV8BbgIGV1mdBryS9ngfyRgffwTsNbN/N7MtZvZvZnY8CRwX7v4q8L9JJfWXgdeBLSRzXAzKNw6m\nUmI+LXeC1xlcwMyOA34ELAyO1BLHzC4H9gRHHoNLZpO6dPYYUv/Y3+Xu55M6il1a3ZCqw8wagL8H\n3gmcARwHXFbNmCIu+zMzYo4td4LfAUxPuz+dzH+BYi84OfRjYEXaV869ZnZK8PippK4xiLv/Clxp\nZi8BK4HZwHLglLQ200iNmbjbDvS5+8bg/o+A9wJ7Ejgu3g886e77gmnc+4CZJHNcDMqXH7LzacF+\nKXeCH7pQKkh0V5M6aZAIZmbAd4Dn3P0raQ8NXhxG8P+fVDq2SnP3xe4+3d3PAv4G6HL3jwMbzGx+\n0CwpfbEdeMXMzg42XQo8T+qzkahxAfwauNjMJgWfl0uBbSRwXKTJlx9+ArQAmNmFwGF37xtpR2Vf\nB29m/41UPfnBC6W+VNYnjBAz+wCpn615lqNfpT4P/ILUCed3kFohcLW7v55zJzFkZrOAG939SjM7\nC/gBqa/mW4GPu/vBqgZYAcFqqm+TOqn4G1IfXCOB48LM2ki9/iPAM8AngNNJwLgws5XALFLfWHYD\nXwDWkGccmNnXgL8ABoBr3X3TiPvXhU4iIvFU9gudRESkOpTgRURiSgleRCSmlOBFRGJKCV5EJKaU\n4EVEYkoJXsYNM3tnelnVrMduNbPZwe2/N7NJedp9y8zOLfF53yw9WpHq0zp4GTeCgm0PBiVkR2r3\nEvC+wYJNITzv79z9+DD2JVJJOoKX8WaCmX0zqMLYbWaTAczse2Z2lZm1kipa9ailfhA+Q/A3Fwa3\n3zSz24MfIHnGzE4Ptp8d3P+lmd2e9fdfCH6I4Xkz+1Kw7a/M7P8Gt0+31I+7xL4SpESfEryMN38E\nfC2owrgb+O/Bdgfc3TtIlZ1tdPcP5fj79K+sxwJPuPsMYD1Hf1C+A/gnd/9TUmVsATCzK4GpwQ8x\nnAf8iZld6u73AzvN7DPA3cAX3D0JhcIk4pTgZbx5afBHVIBfklldr1R/cPe1OfZ1CfDvwe0fprWf\nA8wxs2eC9meTKnML0EqqzlC/u68aQ0wioampdgAiJRpIu32YsR2kpBevOpK1r3wnp25z9+/m2D49\niOcdZmauk1sSATqClzh6C5g8hr9/klRpa4CPpG1fB3zSzOoBzOwdZnZK8FNz3yFVBnkb8LkxPLdI\naHQEL+NN9pFxriPl75A6yfqbPPPwuf7W0+63AqvM7Cagc3C7uz9oZn8MbDKzPwD9wDXADUCPuz9p\nZs8CG83sIXd/odQXJxImLZMUEYkpTdGIiMSUEryISEwpwYuIxJQSvIhITCnBi4jElBK8iEhMKcGL\niMSUEryISEz9f6RVeAOAa3i6AAAAAElFTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x7f40ff659890>"
       ]
      }
     ],
     "prompt_number": 25
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def max_finder (listt) :\n",
      "    m = 0\n",
      "    for i in range (len(listt)) :\n",
      "        if listt[i] > listt[m] :\n",
      "            m = i\n",
      "    return m\n",
      "def compare (out,learn,n = nlayers,m = 4) : # m is ntracks\n",
      "    counter = 0\n",
      "    output = np.zeros(len(out) , dtype=float32)\n",
      "    for i in range (len(out)) :\n",
      "        output[i] = out[i]\n",
      "    learned = learn\n",
      "    for track in range (m) :\n",
      "        subcounter = 0\n",
      "        for layer in range (n) :\n",
      "            start = int(float(layer) * len(output) / n)\n",
      "            end = int(float(layer+1) * len(output) / n)\n",
      "            subOutput = output[start:end]\n",
      "            subLearned = learned[start:end]\n",
      "            if max_finder(subOutput) == max_finder(subLearned) :\n",
      "                subcounter += 1\n",
      "            output[max_finder(subOutput)+start] = 0\n",
      "            learned[max_finder(subLearned)+start] = 0\n",
      "        if subcounter >= 0.75*n :\n",
      "            counter += 1\n",
      "    return counter"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 14
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "counter = np.zeros(5 , dtype=float32)\n",
      "start = time.mktime(time.gmtime())\n",
      "for i in range (200) :\n",
      "    counter[compare(ds_output[cut+2000+i],trainer.network.predict(valid[0])[2000+i],4,4)] += 1\n",
      "stop = time.mktime (time.gmtime())\n",
      "spent.append (stop - start)\n",
      "print \"\\t\" , stop - start\n",
      "print counter"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\t384.0\n",
        "[ 45.  62.  67.  11.  15.]\n"
       ]
      }
     ],
     "prompt_number": 16
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "gots = np.zeros(5 , dtype=float32)\n",
      "start = time.mktime(time.gmtime())\n",
      "for i in range (200) :\n",
      "    n_goodTracks = int(4*got(ds_output[cut+i],trainer.network.predict(valid[0])[i],0.1,0.75))\n",
      "    if n_goodTracks == 4 :\n",
      "        print \"\\t\",i\n",
      "    gots[n_goodTracks] += 1\n",
      "stop = time.mktime (time.gmtime())\n",
      "spent.append (stop - start)\n",
      "print \"\\t\" , stop - start\n",
      "print \"gots :\\t\",gots"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\t30\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "46\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "47\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "66\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "79\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "80\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "144\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "160\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "163\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "166\n",
        "\t"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "451.0\n",
        "gots :\t[ 18.  84.  55.  33.  10.]\n"
       ]
      }
     ],
     "prompt_number": 10
    }
   ],
   "metadata": {}
  }
 ]
}