MergeNetThreeLegDropout_image_prob_mergedarray_LARGE.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 0: GeForce GTX TITAN (CNMeM is disabled, CuDNN 4007)\n",
      "/home/d/nolearn/local/lib/python2.7/site-packages/theano/tensor/signal/downsample.py:6: UserWarning: downsample module has been moved to the theano.tensor.signal.pool module.\n",
      "  \"downsample module has been moved to the theano.tensor.signal.pool module.\")\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import mlproof as mlp\n",
    "import mlproof.nets as nets\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "from nolearn.lasagne.visualize import plot_loss\n",
    "from nolearn.lasagne.visualize import plot_conv_weights\n",
    "from nolearn.lasagne.visualize import plot_conv_activity\n",
    "from nolearn.lasagne.visualize import plot_occlusion\n",
    "\n",
    "from matplotlib.pyplot import imshow\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CNN configuration: \n",
      "    Our CNN with different \"legs\" for image, prob, merged_array.\n",
      "\n",
      "    Now with dropout layers. Three leg version!\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "cnn = nets.MergeNetThreeLegDropout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded /home/d/patches//cylinder1/ in 0.168568849564 seconds.\n"
     ]
    }
   ],
   "source": [
    "X_train, y_train, X_test, y_test = mlp.Patch.load('cylinder1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# Neural Network with 1649348 learnable parameters\n",
      "\n",
      "## Layer information\n",
      "\n",
      "  #  name            size\n",
      "---  --------------  --------\n",
      "  0  image_input     1x75x75\n",
      "  1  image_conv1     24x63x63\n",
      "  2  image_pool1     24x31x31\n",
      "  3  image_conv2     22x19x19\n",
      "  4  image_pool2     22x9x9\n",
      "  5  image_dropout   22x9x9\n",
      "  6  prob_input      1x75x75\n",
      "  7  prob_conv1      24x63x63\n",
      "  8  prob_pool1      24x31x31\n",
      "  9  prob_conv2      22x19x19\n",
      " 10  prob_pool2      22x9x9\n",
      " 11  prob_dropout    22x9x9\n",
      " 12  binary_input    1x75x75\n",
      " 13  binary_conv1    24x63x63\n",
      " 14  binary_pool1    24x31x31\n",
      " 15  binary_conv2    22x19x19\n",
      " 16  binary_pool2    22x9x9\n",
      " 17  binary_dropout  22x9x9\n",
      " 18  merge           66x9x9\n",
      " 19  hidden3         256\n",
      " 20  dropout3        256\n",
      " 21  output          2\n",
      "\n",
      "  epoch    train loss    valid loss    train/val    valid acc  dur\n",
      "-------  ------------  ------------  -----------  -----------  -------\n",
      "      1       \u001b[36m0.35306\u001b[0m       \u001b[32m0.29950\u001b[0m      1.17884      0.87800  669.94s\n",
      "      2       \u001b[36m0.31101\u001b[0m       \u001b[32m0.28596\u001b[0m      1.08761      0.88519  675.25s\n",
      "      3       \u001b[36m0.29858\u001b[0m       \u001b[32m0.27550\u001b[0m      1.08379      0.88866  657.67s\n",
      "      4       \u001b[36m0.28967\u001b[0m       \u001b[32m0.26866\u001b[0m      1.07821      0.89161  691.34s\n",
      "      5       \u001b[36m0.28228\u001b[0m       \u001b[32m0.26356\u001b[0m      1.07106      0.89392  707.98s\n",
      "      6       \u001b[36m0.27745\u001b[0m       \u001b[32m0.25886\u001b[0m      1.07184      0.89690  709.32s\n",
      "      7       \u001b[36m0.27156\u001b[0m       \u001b[32m0.25426\u001b[0m      1.06803      0.89918  723.68s\n",
      "      8       \u001b[36m0.26743\u001b[0m       \u001b[32m0.25233\u001b[0m      1.05984      0.89882  679.94s\n",
      "      9       \u001b[36m0.26307\u001b[0m       \u001b[32m0.24754\u001b[0m      1.06271      0.90206  673.96s\n",
      "     10       \u001b[36m0.25854\u001b[0m       \u001b[32m0.24336\u001b[0m      1.06241      0.90389  670.98s\n",
      "     11       \u001b[36m0.25633\u001b[0m       \u001b[32m0.23950\u001b[0m      1.07028      0.90492  677.60s\n",
      "     12       \u001b[36m0.25253\u001b[0m       \u001b[32m0.23792\u001b[0m      1.06144      0.90599  711.58s\n",
      "     13       \u001b[36m0.24979\u001b[0m       \u001b[32m0.23553\u001b[0m      1.06054      0.90680  702.72s\n",
      "     14       \u001b[36m0.24666\u001b[0m       \u001b[32m0.23080\u001b[0m      1.06871      0.90878  653.86s\n",
      "     15       \u001b[36m0.24447\u001b[0m       \u001b[32m0.23008\u001b[0m      1.06255      0.90918  657.93s\n",
      "     16       \u001b[36m0.24171\u001b[0m       \u001b[32m0.22741\u001b[0m      1.06287      0.90909  692.49s\n",
      "     17       \u001b[36m0.23983\u001b[0m       \u001b[32m0.22420\u001b[0m      1.06973      0.91101  716.97s\n",
      "     18       \u001b[36m0.23718\u001b[0m       \u001b[32m0.22265\u001b[0m      1.06524      0.91254  700.06s\n",
      "     19       \u001b[36m0.23375\u001b[0m       \u001b[32m0.22061\u001b[0m      1.05957      0.91281  702.96s\n",
      "     20       \u001b[36m0.23142\u001b[0m       \u001b[32m0.21873\u001b[0m      1.05800      0.91384  654.36s\n",
      "     21       \u001b[36m0.22904\u001b[0m       \u001b[32m0.21860\u001b[0m      1.04776      0.91393  697.39s\n",
      "     22       \u001b[36m0.22689\u001b[0m       \u001b[32m0.21378\u001b[0m      1.06134      0.91504  702.25s\n",
      "     23       \u001b[36m0.22406\u001b[0m       \u001b[32m0.21143\u001b[0m      1.05974      0.91636  702.82s\n",
      "     24       \u001b[36m0.22249\u001b[0m       \u001b[32m0.21050\u001b[0m      1.05696      0.91663  668.09s\n",
      "     25       \u001b[36m0.21982\u001b[0m       \u001b[32m0.20841\u001b[0m      1.05475      0.91817  685.05s\n",
      "     26       \u001b[36m0.21809\u001b[0m       \u001b[32m0.20679\u001b[0m      1.05466      0.91796  671.49s\n",
      "     27       \u001b[36m0.21559\u001b[0m       0.20692      1.04195      0.91790  644.72s\n",
      "     28       \u001b[36m0.21351\u001b[0m       \u001b[32m0.20190\u001b[0m      1.05752      0.91960  643.22s\n",
      "     29       \u001b[36m0.21070\u001b[0m       \u001b[32m0.19995\u001b[0m      1.05377      0.92076  642.11s\n",
      "     30       \u001b[36m0.20972\u001b[0m       0.20221      1.03713      0.92009  649.72s\n",
      "     31       \u001b[36m0.20775\u001b[0m       \u001b[32m0.19835\u001b[0m      1.04742      0.92213  641.64s\n",
      "     32       \u001b[36m0.20574\u001b[0m       \u001b[32m0.19268\u001b[0m      1.06777      0.92457  642.99s\n",
      "     33       \u001b[36m0.20356\u001b[0m       \u001b[32m0.19196\u001b[0m      1.06045      0.92575  641.55s\n",
      "     34       \u001b[36m0.20285\u001b[0m       \u001b[32m0.19059\u001b[0m      1.06435      0.92537  639.19s\n",
      "     35       \u001b[36m0.19879\u001b[0m       \u001b[32m0.18869\u001b[0m      1.05350      0.92670  699.72s\n",
      "     36       \u001b[36m0.19713\u001b[0m       \u001b[32m0.18818\u001b[0m      1.04757      0.92635  662.45s\n",
      "     37       \u001b[36m0.19490\u001b[0m       \u001b[32m0.18490\u001b[0m      1.05408      0.92709  638.74s\n",
      "     38       0.19496       0.18518      1.05286      0.92723  670.11s\n",
      "     39       \u001b[36m0.19285\u001b[0m       \u001b[32m0.18329\u001b[0m      1.05217      0.92792  638.73s\n",
      "     40       \u001b[36m0.19128\u001b[0m       \u001b[32m0.18318\u001b[0m      1.04423      0.92846  639.04s\n",
      "     41       \u001b[36m0.18772\u001b[0m       \u001b[32m0.17901\u001b[0m      1.04864      0.92944  639.09s\n",
      "     42       \u001b[36m0.18683\u001b[0m       \u001b[32m0.17771\u001b[0m      1.05133      0.93042  636.65s\n",
      "     43       \u001b[36m0.18569\u001b[0m       \u001b[32m0.17415\u001b[0m      1.06622      0.93167  637.26s\n",
      "     44       \u001b[36m0.18316\u001b[0m       0.17515      1.04572      0.93153  631.70s\n",
      "     45       \u001b[36m0.18173\u001b[0m       \u001b[32m0.17127\u001b[0m      1.06105      0.93336  636.05s\n",
      "     46       \u001b[36m0.18039\u001b[0m       0.17200      1.04883      0.93339  631.06s\n",
      "     47       \u001b[36m0.17912\u001b[0m       \u001b[32m0.16952\u001b[0m      1.05662      0.93382  631.29s\n",
      "     48       \u001b[36m0.17726\u001b[0m       0.16980      1.04394      0.93332  630.53s\n",
      "     49       \u001b[36m0.17696\u001b[0m       \u001b[32m0.16410\u001b[0m      1.07837      0.93674  630.68s\n",
      "     50       \u001b[36m0.17506\u001b[0m       0.16473      1.06272      0.93603  629.87s\n",
      "     51       \u001b[36m0.17220\u001b[0m       \u001b[32m0.16331\u001b[0m      1.05443      0.93653  630.93s\n",
      "     52       0.17225       \u001b[32m0.16219\u001b[0m      1.06202      0.93779  627.18s\n",
      "     53       \u001b[36m0.16943\u001b[0m       \u001b[32m0.16044\u001b[0m      1.05600      0.93804  624.57s\n",
      "     54       \u001b[36m0.16871\u001b[0m       0.16605      1.01605      0.93491  621.48s\n",
      "     55       \u001b[36m0.16807\u001b[0m       \u001b[32m0.15975\u001b[0m      1.05207      0.93840  624.34s\n",
      "     56       \u001b[36m0.16757\u001b[0m       \u001b[32m0.15603\u001b[0m      1.07395      0.94058  623.73s\n",
      "     57       \u001b[36m0.16548\u001b[0m       \u001b[32m0.15386\u001b[0m      1.07553      0.94050  623.73s\n",
      "     58       \u001b[36m0.16440\u001b[0m       0.15422      1.06603      0.94076  618.55s\n",
      "     59       \u001b[36m0.16300\u001b[0m       0.15499      1.05167      0.94102  617.47s\n",
      "     60       \u001b[36m0.16135\u001b[0m       \u001b[32m0.14924\u001b[0m      1.08119      0.94288  617.60s\n",
      "     61       \u001b[36m0.16079\u001b[0m       \u001b[32m0.14912\u001b[0m      1.07824      0.94259  617.09s\n",
      "     62       \u001b[36m0.15994\u001b[0m       \u001b[32m0.14558\u001b[0m      1.09864      0.94546  617.33s\n",
      "     63       \u001b[36m0.15918\u001b[0m       0.14863      1.07102      0.94244  617.71s\n",
      "     64       \u001b[36m0.15636\u001b[0m       0.14742      1.06063      0.94350  617.90s\n",
      "     65       \u001b[36m0.15605\u001b[0m       0.14613      1.06784      0.94376  617.89s\n",
      "     66       \u001b[36m0.15408\u001b[0m       \u001b[32m0.14457\u001b[0m      1.06577      0.94498  617.82s\n",
      "     67       0.15431       0.14733      1.04740      0.94241  617.43s\n",
      "     68       \u001b[36m0.15402\u001b[0m       \u001b[32m0.14309\u001b[0m      1.07636      0.94474  616.94s\n",
      "     69       \u001b[36m0.15214\u001b[0m       \u001b[32m0.14167\u001b[0m      1.07396      0.94546  617.95s\n",
      "     70       \u001b[36m0.15140\u001b[0m       \u001b[32m0.13967\u001b[0m      1.08396      0.94681  617.39s\n",
      "     71       \u001b[36m0.14995\u001b[0m       0.14010      1.07031      0.94694  618.09s\n",
      "     72       \u001b[36m0.14875\u001b[0m       \u001b[32m0.13899\u001b[0m      1.07027      0.94743  618.22s\n",
      "     73       0.14883       \u001b[32m0.13720\u001b[0m      1.08472      0.94815  618.13s\n",
      "     74       \u001b[36m0.14742\u001b[0m       0.13729      1.07376      0.94798  621.49s\n",
      "     75       \u001b[36m0.14628\u001b[0m       \u001b[32m0.13591\u001b[0m      1.07634      0.94881  621.49s\n",
      "     76       \u001b[36m0.14555\u001b[0m       \u001b[32m0.13203\u001b[0m      1.10238      0.95102  621.54s\n",
      "     77       \u001b[36m0.14539\u001b[0m       0.13451      1.08087      0.94935  621.50s\n",
      "     78       \u001b[36m0.14378\u001b[0m       0.13467      1.06767      0.94918  621.31s\n",
      "     79       \u001b[36m0.14335\u001b[0m       0.13228      1.08370      0.95027  621.69s\n",
      "     80       \u001b[36m0.14334\u001b[0m       0.13308      1.07713      0.94981  621.55s\n",
      "     81       \u001b[36m0.14155\u001b[0m       0.13286      1.06539      0.94957  618.27s\n",
      "     82       0.14161       \u001b[32m0.13176\u001b[0m      1.07483      0.95039  645.77s\n",
      "     83       \u001b[36m0.13965\u001b[0m       \u001b[32m0.12886\u001b[0m      1.08372      0.95145  637.70s\n",
      "     84       \u001b[36m0.13923\u001b[0m       \u001b[32m0.12852\u001b[0m      1.08340      0.95161  617.81s\n",
      "     85       0.13934       \u001b[32m0.12769\u001b[0m      1.09127      0.95150  617.83s\n",
      "     86       \u001b[36m0.13849\u001b[0m       \u001b[32m0.12746\u001b[0m      1.08649      0.95191  617.76s\n",
      "     87       \u001b[36m0.13834\u001b[0m       0.12750      1.08502      0.95217  617.63s\n",
      "     88       \u001b[36m0.13746\u001b[0m       0.12821      1.07214      0.95174  617.57s\n",
      "     89       \u001b[36m0.13659\u001b[0m       0.12796      1.06742      0.95139  617.65s\n",
      "     90       \u001b[36m0.13631\u001b[0m       \u001b[32m0.12339\u001b[0m      1.10473      0.95322  617.44s\n",
      "     91       \u001b[36m0.13531\u001b[0m       \u001b[32m0.12243\u001b[0m      1.10525      0.95379  617.79s\n",
      "     92       \u001b[36m0.13483\u001b[0m       \u001b[32m0.12118\u001b[0m      1.11265      0.95482  618.74s\n",
      "     93       0.13487       0.12274      1.09879      0.95403  617.46s\n",
      "     94       \u001b[36m0.13397\u001b[0m       \u001b[32m0.12079\u001b[0m      1.10912      0.95573  617.89s\n",
      "     95       \u001b[36m0.13315\u001b[0m       \u001b[32m0.11752\u001b[0m      1.13297      0.95609  617.20s\n",
      "     96       \u001b[36m0.13270\u001b[0m       0.11818      1.12285      0.95666  617.84s\n",
      "     97       \u001b[36m0.13201\u001b[0m       \u001b[32m0.11525\u001b[0m      1.14541      0.95743  617.72s\n",
      "     98       \u001b[36m0.13008\u001b[0m       0.11990      1.08489      0.95444  617.20s\n",
      "     99       \u001b[36m0.12988\u001b[0m       0.11549      1.12464      0.95754  617.56s\n",
      "    100       \u001b[36m0.12894\u001b[0m       \u001b[32m0.11383\u001b[0m      1.13276      0.95794  618.05s\n",
      "    101       \u001b[36m0.12874\u001b[0m       0.11525      1.11710      0.95736  618.09s\n",
      "    102       \u001b[36m0.12748\u001b[0m       0.11597      1.09925      0.95699  617.88s\n",
      "    103       0.12775       0.11624      1.09903      0.95706  617.69s\n",
      "    104       \u001b[36m0.12734\u001b[0m       \u001b[32m0.11334\u001b[0m      1.12352      0.95832  618.16s\n",
      "    105       0.12769       0.11625      1.09842      0.95633  617.50s\n",
      "    106       \u001b[36m0.12654\u001b[0m       0.11438      1.10631      0.95783  617.44s\n",
      "    107       \u001b[36m0.12574\u001b[0m       \u001b[32m0.11328\u001b[0m      1.11003      0.95790  617.32s\n",
      "    108       \u001b[36m0.12499\u001b[0m       0.11485      1.08830      0.95701  618.30s\n",
      "    109       0.12503       0.11718      1.06698      0.95661  618.14s\n",
      "    110       \u001b[36m0.12428\u001b[0m       \u001b[32m0.11223\u001b[0m      1.10732      0.95924  617.95s\n",
      "    111       \u001b[36m0.12413\u001b[0m       \u001b[32m0.11145\u001b[0m      1.11376      0.95894  618.90s\n",
      "    112       \u001b[36m0.12351\u001b[0m       0.11283      1.09468      0.95859  617.96s\n",
      "    113       0.12402       \u001b[32m0.10740\u001b[0m      1.15475      0.96093  618.70s\n",
      "    114       \u001b[36m0.12244\u001b[0m       0.11150      1.09812      0.95885  617.54s\n",
      "    115       0.12322       0.11087      1.11140      0.95872  617.37s\n",
      "    116       \u001b[36m0.12208\u001b[0m       0.10955      1.11438      0.96001  617.54s\n",
      "    117       0.12229       0.11081      1.10360      0.95932  622.59s\n",
      "    118       \u001b[36m0.12090\u001b[0m       \u001b[32m0.10625\u001b[0m      1.13785      0.96158  622.86s\n",
      "    119       \u001b[36m0.11964\u001b[0m       0.10658      1.12250      0.96083  623.37s\n",
      "    120       \u001b[36m0.11915\u001b[0m       \u001b[32m0.10594\u001b[0m      1.12464      0.96172  626.87s\n",
      "    121       0.11921       0.10683      1.11589      0.96100  632.63s\n",
      "    122       \u001b[36m0.11841\u001b[0m       0.10864      1.08992      0.96088  632.79s\n",
      "    123       \u001b[36m0.11805\u001b[0m       \u001b[32m0.10314\u001b[0m      1.14452      0.96308  631.20s\n",
      "    124       0.11836       \u001b[32m0.10234\u001b[0m      1.15658      0.96285  631.23s\n",
      "    125       0.11820       0.10514      1.12427      0.96189  631.54s\n",
      "    126       \u001b[36m0.11647\u001b[0m       0.10545      1.10456      0.96148  631.34s\n",
      "    127       0.11680       0.10569      1.10520      0.96151  631.34s\n",
      "    128       \u001b[36m0.11630\u001b[0m       \u001b[32m0.10066\u001b[0m      1.15539      0.96412  631.46s\n",
      "    129       0.11630       0.10514      1.10616      0.96177  630.69s\n",
      "    130       \u001b[36m0.11584\u001b[0m       \u001b[32m0.09946\u001b[0m      1.16471      0.96448  629.81s\n",
      "    131       \u001b[36m0.11490\u001b[0m       0.09963      1.15318      0.96517  624.98s\n",
      "    132       0.11542       0.10268      1.12407      0.96294  625.00s\n",
      "    133       \u001b[36m0.11456\u001b[0m       \u001b[32m0.09927\u001b[0m      1.15403      0.96481  624.72s\n",
      "    134       \u001b[36m0.11436\u001b[0m       \u001b[32m0.09920\u001b[0m      1.15281      0.96470  629.74s\n",
      "    135       0.11502       0.09986      1.15181      0.96386  627.97s\n",
      "    136       \u001b[36m0.11290\u001b[0m       \u001b[32m0.09908\u001b[0m      1.13957      0.96437  626.71s\n",
      "    137       \u001b[36m0.11280\u001b[0m       \u001b[32m0.09825\u001b[0m      1.14818      0.96493  628.21s\n",
      "    138       \u001b[36m0.11238\u001b[0m       0.09861      1.13970      0.96517  630.77s\n",
      "    139       0.11278       0.09973      1.13079      0.96423  630.28s\n",
      "    140       0.11304       0.10029      1.12714      0.96417  630.65s\n",
      "    141       \u001b[36m0.11161\u001b[0m       0.09962      1.12036      0.96420  631.75s\n",
      "    142       \u001b[36m0.11110\u001b[0m       0.10145      1.09510      0.96353  631.21s\n",
      "    143       0.11225       \u001b[32m0.09735\u001b[0m      1.15311      0.96550  631.44s\n",
      "    144       \u001b[36m0.11100\u001b[0m       \u001b[32m0.09419\u001b[0m      1.17848      0.96757  631.92s\n",
      "    145       0.11101       0.09676      1.14722      0.96574  631.55s\n",
      "    146       0.11126       0.09787      1.13683      0.96549  631.46s\n",
      "    147       \u001b[36m0.10974\u001b[0m       0.09705      1.13081      0.96577  637.48s\n",
      "    148       0.10981       0.09467      1.15997      0.96685  641.23s\n",
      "    149       \u001b[36m0.10918\u001b[0m       0.09640      1.13262      0.96601  636.21s\n",
      "    150       \u001b[36m0.10904\u001b[0m       0.09456      1.15314      0.96632  634.99s\n",
      "    151       \u001b[36m0.10847\u001b[0m       0.09518      1.13964      0.96594  635.97s\n",
      "    152       \u001b[36m0.10843\u001b[0m       \u001b[32m0.09255\u001b[0m      1.17160      0.96769  635.63s\n",
      "    153       \u001b[36m0.10769\u001b[0m       0.09349      1.15186      0.96773  631.78s\n",
      "    154       0.10864       0.09351      1.16187      0.96725  630.24s\n",
      "    155       \u001b[36m0.10729\u001b[0m       0.09338      1.14900      0.96669  626.33s\n",
      "    156       \u001b[36m0.10716\u001b[0m       0.09293      1.15308      0.96744  624.81s\n",
      "    157       \u001b[36m0.10639\u001b[0m       \u001b[32m0.09231\u001b[0m      1.15243      0.96752  626.62s\n",
      "    158       0.10668       0.09566      1.11521      0.96587  627.19s\n",
      "    159       0.10639       0.09522      1.11737      0.96624  630.40s\n",
      "    160       \u001b[36m0.10628\u001b[0m       0.09446      1.12513      0.96715  636.81s\n",
      "    161       \u001b[36m0.10564\u001b[0m       \u001b[32m0.08934\u001b[0m      1.18242      0.96883  629.00s\n",
      "    162       0.10637       0.09275      1.14677      0.96740  638.68s\n",
      "    163       0.10594       0.08935      1.18566      0.96840  639.64s\n",
      "    164       \u001b[36m0.10410\u001b[0m       0.09350      1.11339      0.96748  638.59s\n",
      "    165       0.10507       0.09433      1.11387      0.96680  638.68s\n",
      "    166       0.10457       0.09184      1.13856      0.96785  638.24s\n",
      "    167       0.10477       0.09259      1.13155      0.96783  640.07s\n",
      "    168       \u001b[36m0.10388\u001b[0m       0.09225      1.12608      0.96765  639.58s\n",
      "    169       \u001b[36m0.10318\u001b[0m       \u001b[32m0.08877\u001b[0m      1.16238      0.96910  639.33s\n",
      "    170       0.10394       0.09244      1.12445      0.96772  638.02s\n",
      "    171       \u001b[36m0.10314\u001b[0m       \u001b[32m0.08763\u001b[0m      1.17704      0.96942  639.14s\n",
      "    172       0.10383       0.09160      1.13350      0.96816  638.54s\n",
      "    173       \u001b[36m0.10225\u001b[0m       0.08937      1.14417      0.96896  639.39s\n",
      "    174       0.10275       0.08993      1.14256      0.96897  637.92s\n",
      "    175       \u001b[36m0.10194\u001b[0m       0.08980      1.13507      0.96833  638.63s\n",
      "    176       0.10228       \u001b[32m0.08568\u001b[0m      1.19379      0.97032  638.38s\n",
      "    177       0.10224       0.08750      1.16849      0.96994  634.89s\n",
      "    178       \u001b[36m0.10171\u001b[0m       0.08976      1.13318      0.96925  623.79s\n",
      "    179       0.10216       0.08646      1.18157      0.97064  622.69s\n",
      "    180       \u001b[36m0.10136\u001b[0m       0.08865      1.14335      0.96955  621.75s\n",
      "    181       \u001b[36m0.10098\u001b[0m       \u001b[32m0.08516\u001b[0m      1.18577      0.97189  620.23s\n",
      "    182       \u001b[36m0.10085\u001b[0m       0.08695      1.15986      0.97049  623.54s\n",
      "    183       \u001b[36m0.10018\u001b[0m       0.08822      1.13552      0.96980  621.19s\n",
      "    184       \u001b[36m0.09972\u001b[0m       0.08558      1.16524      0.97173  618.97s\n",
      "    185       0.10058       0.08812      1.14130      0.96915  619.79s\n",
      "    186       0.10018       0.08612      1.16329      0.97133  622.13s\n",
      "    187       0.10092       0.08579      1.17634      0.97093  618.96s\n",
      "    188       \u001b[36m0.09891\u001b[0m       0.08659      1.14232      0.97007  622.17s\n",
      "    189       0.09925       \u001b[32m0.08382\u001b[0m      1.18407      0.97189  623.01s\n",
      "    190       0.09947       0.08556      1.16260      0.97104  618.54s\n",
      "    191       0.09925       0.08540      1.16206      0.97096  623.09s\n",
      "    192       \u001b[36m0.09868\u001b[0m       0.08413      1.17292      0.97177  618.90s\n",
      "    193       \u001b[36m0.09835\u001b[0m       0.08466      1.16180      0.97106  618.89s\n",
      "    194       \u001b[36m0.09830\u001b[0m       0.08385      1.17233      0.97130  619.71s\n",
      "    195       \u001b[36m0.09756\u001b[0m       0.08499      1.14793      0.97074  620.12s\n",
      "    196       0.09843       0.08572      1.14830      0.97099  618.40s\n",
      "    197       0.09802       0.08424      1.16356      0.97125  618.55s\n",
      "    198       0.09803       0.08439      1.16154      0.97131  618.22s\n",
      "    199       \u001b[36m0.09717\u001b[0m       0.08750      1.11051      0.97061  618.25s\n",
      "    200       \u001b[36m0.09693\u001b[0m       \u001b[32m0.08241\u001b[0m      1.17618      0.97264  618.21s\n",
      "    201       0.09695       0.08458      1.14625      0.97099  618.48s\n",
      "    202       \u001b[36m0.09638\u001b[0m       0.08499      1.13403      0.97171  618.75s\n",
      "    203       0.09687       0.08323      1.16389      0.97195  618.54s\n",
      "    204       \u001b[36m0.09634\u001b[0m       0.08396      1.14744      0.97168  618.65s\n",
      "    205       0.09688       \u001b[32m0.08079\u001b[0m      1.19920      0.97335  618.46s\n",
      "    206       0.09672       0.08385      1.15354      0.97227  614.54s\n",
      "    207       0.09727       0.08376      1.16130      0.97279  617.97s\n",
      "    208       \u001b[36m0.09571\u001b[0m       \u001b[32m0.08022\u001b[0m      1.19309      0.97316  617.84s\n",
      "    209       0.09592       0.08210      1.16827      0.97306  618.56s\n",
      "    210       \u001b[36m0.09544\u001b[0m       0.08397      1.13670      0.97156  618.79s\n",
      "    211       \u001b[36m0.09488\u001b[0m       \u001b[32m0.07970\u001b[0m      1.19054      0.97418  618.63s\n",
      "    212       0.09620       0.08263      1.16429      0.97265  614.58s\n",
      "    213       0.09520       0.08277      1.15017      0.97258  617.88s\n",
      "    214       0.09539       0.08304      1.14869      0.97215  617.59s\n",
      "    215       \u001b[36m0.09383\u001b[0m       0.08246      1.13779      0.97267  617.84s\n",
      "    216       0.09427       \u001b[32m0.07901\u001b[0m      1.19315      0.97446  618.26s\n",
      "    217       0.09501       0.07946      1.19567      0.97469  618.80s\n",
      "    218       0.09396       0.08332      1.12777      0.97222  618.84s\n",
      "    219       0.09412       0.08099      1.16207      0.97327  618.91s\n",
      "    220       \u001b[36m0.09334\u001b[0m       0.07979      1.16973      0.97394  619.04s\n",
      "    221       0.09355       \u001b[32m0.07846\u001b[0m      1.19240      0.97405  618.80s\n",
      "    222       \u001b[36m0.09324\u001b[0m       \u001b[32m0.07638\u001b[0m      1.22068      0.97501  614.56s\n",
      "    223       \u001b[36m0.09310\u001b[0m       0.08018      1.16109      0.97354  618.66s\n",
      "    224       \u001b[36m0.09211\u001b[0m       0.08007      1.15040      0.97370  618.80s\n",
      "    225       0.09304       0.07940      1.17181      0.97346  618.84s\n",
      "    226       0.09409       0.07775      1.21012      0.97481  619.03s\n",
      "    227       0.09248       0.07765      1.19091      0.97481  619.11s\n",
      "    228       0.09231       0.07918      1.16595      0.97389  618.83s\n",
      "    229       \u001b[36m0.09156\u001b[0m       0.07925      1.15527      0.97407  618.56s\n",
      "    230       \u001b[36m0.09129\u001b[0m       0.07653      1.19285      0.97604  618.76s\n",
      "    231       \u001b[36m0.09112\u001b[0m       0.07839      1.16235      0.97429  619.03s\n",
      "    232       0.09203       0.08071      1.14030      0.97341  619.29s\n",
      "    233       0.09200       0.07942      1.15844      0.97374  618.63s\n",
      "    234       \u001b[36m0.09095\u001b[0m       0.07698      1.18161      0.97445  619.01s\n",
      "    235       0.09124       0.07992      1.14165      0.97339  618.73s\n",
      "    236       0.09161       \u001b[32m0.07621\u001b[0m      1.20204      0.97585  618.94s\n",
      "    237       0.09113       0.07907      1.15251      0.97379  618.42s\n",
      "    238       0.09101       0.08045      1.13124      0.97365  618.39s\n",
      "    239       \u001b[36m0.09091\u001b[0m       0.07967      1.14097      0.97430  618.24s\n",
      "    240       0.09146       0.07858      1.16389      0.97407  618.31s\n",
      "    241       0.09133       0.08089      1.12902      0.97303  618.21s\n",
      "    242       0.09114       0.07914      1.15158      0.97464  618.11s\n",
      "    243       \u001b[36m0.09002\u001b[0m       0.07837      1.14867      0.97507  618.50s\n",
      "    244       \u001b[36m0.08975\u001b[0m       0.08196      1.09498      0.97272  618.93s\n",
      "    245       \u001b[36m0.08927\u001b[0m       \u001b[32m0.07501\u001b[0m      1.19023      0.97634  622.82s\n",
      "    246       0.08933       0.07908      1.12969      0.97437  620.97s\n",
      "    247       0.08984       0.07636      1.17659      0.97556  623.91s\n",
      "    248       \u001b[36m0.08889\u001b[0m       0.07732      1.14974      0.97538  624.99s\n",
      "    249       \u001b[36m0.08859\u001b[0m       \u001b[32m0.07415\u001b[0m      1.19461      0.97683  625.16s\n",
      "    250       \u001b[36m0.08846\u001b[0m       0.07782      1.13677      0.97442  625.13s\n",
      "    251       0.08860       0.07881      1.12419      0.97508  625.09s\n",
      "    252       \u001b[36m0.08789\u001b[0m       0.07816      1.12447      0.97424  630.04s\n",
      "    253       0.08990       0.07503      1.19824      0.97649  631.86s\n",
      "    254       0.08802       0.07467      1.17870      0.97585  632.68s\n",
      "    255       0.08873       0.07677      1.15580      0.97539  636.58s\n",
      "    256       0.08791       0.07602      1.15647      0.97596  632.91s\n",
      "    257       0.08865       0.07473      1.18631      0.97682  633.90s\n",
      "    258       0.08822       0.07507      1.17511      0.97613  638.80s\n",
      "    259       0.08858       0.07852      1.12813      0.97517  639.53s\n",
      "    260       0.08815       0.08059      1.09380      0.97394  639.21s\n",
      "    261       0.08837       0.07867      1.12338      0.97482  636.77s\n",
      "    262       \u001b[36m0.08702\u001b[0m       0.07453      1.16766      0.97625  639.30s\n",
      "    263       0.08790       0.07610      1.15503      0.97554  639.76s\n",
      "    264       0.08707       0.07511      1.15915      0.97650  641.61s\n",
      "    265       0.08791       0.08455      1.03975      0.97234  640.29s\n",
      "    266       0.08808       0.07640      1.15285      0.97565  644.99s\n",
      "    267       \u001b[36m0.08668\u001b[0m       0.07437      1.16547      0.97646  647.06s\n",
      "    268       0.08693       0.07547      1.15186      0.97607  643.40s\n",
      "    269       \u001b[36m0.08628\u001b[0m       \u001b[32m0.07355\u001b[0m      1.17312      0.97651  646.23s\n",
      "    270       \u001b[36m0.08600\u001b[0m       0.07527      1.14258      0.97642  647.16s\n",
      "    271       0.08691       0.07546      1.15169      0.97618  642.66s\n",
      "    272       0.08614       0.07528      1.14436      0.97598  646.81s\n",
      "    273       0.08725       \u001b[32m0.07274\u001b[0m      1.19940      0.97775  640.47s\n",
      "    274       0.08601       0.07640      1.12583      0.97553  640.34s\n",
      "    275       \u001b[36m0.08588\u001b[0m       \u001b[32m0.07255\u001b[0m      1.18380      0.97759  640.33s\n",
      "    276       \u001b[36m0.08511\u001b[0m       0.07505      1.13402      0.97605  643.20s\n",
      "    277       0.08523       0.07361      1.15782      0.97681  640.11s\n",
      "    278       0.08584       0.07341      1.16939      0.97731  640.13s\n",
      "    279       0.08651       0.07512      1.15157      0.97635  640.59s\n",
      "    280       0.08618       0.07342      1.17386      0.97692  639.15s\n",
      "    281       0.08519       0.07548      1.12861      0.97588  640.07s\n",
      "    282       0.08539       0.07504      1.13791      0.97629  639.18s\n",
      "    283       0.08523       0.07570      1.12587      0.97643  639.44s\n",
      "    284       \u001b[36m0.08459\u001b[0m       0.07372      1.14746      0.97689  639.20s\n",
      "    285       \u001b[36m0.08398\u001b[0m       0.07298      1.15063      0.97731  638.23s\n",
      "    286       0.08506       0.07405      1.14858      0.97658  638.08s\n",
      "    287       0.08468       0.07510      1.12746      0.97689  638.80s\n",
      "    288       0.08493       0.07357      1.15439      0.97683  639.08s\n",
      "    289       0.08446       \u001b[32m0.07134\u001b[0m      1.18397      0.97783  637.27s\n",
      "    290       \u001b[36m0.08345\u001b[0m       0.07257      1.14997      0.97773  636.76s\n",
      "    291       0.08496       0.07589      1.11950      0.97663  637.37s\n",
      "    292       0.08412       0.07192      1.16969      0.97802  638.91s\n",
      "    293       0.08420       0.07309      1.15208      0.97768  637.77s\n",
      "    294       0.08441       0.07416      1.13825      0.97729  638.18s\n",
      "    295       0.08420       0.07371      1.14230      0.97806  638.72s\n",
      "    296       0.08391       0.07170      1.17017      0.97816  639.18s\n",
      "    297       0.08436       0.07152      1.17952      0.97864  637.82s\n",
      "    298       0.08410       0.07170      1.17297      0.97827  637.66s\n",
      "    299       \u001b[36m0.08304\u001b[0m       0.07543      1.10089      0.97667  638.83s\n",
      "    300       0.08392       \u001b[32m0.07037\u001b[0m      1.19249      0.97886  638.97s\n",
      "    301       0.08399       0.07232      1.16140      0.97803  637.73s\n",
      "    302       0.08323       0.07346      1.13300      0.97805  632.56s\n",
      "    303       \u001b[36m0.08291\u001b[0m       0.07584      1.09309      0.97641  638.28s\n",
      "    304       \u001b[36m0.08233\u001b[0m       0.07066      1.16513      0.97873  632.51s\n",
      "    305       0.08349       0.07361      1.13432      0.97774  632.52s\n",
      "    306       0.08293       0.07445      1.11387      0.97715  632.22s\n",
      "    307       0.08273       \u001b[32m0.06966\u001b[0m      1.18770      0.97912  632.04s\n",
      "    308       \u001b[36m0.08211\u001b[0m       0.07612      1.07860      0.97638  628.02s\n",
      "    309       0.08221       0.07453      1.10297      0.97733  630.88s\n",
      "    310       0.08298       0.07503      1.10596      0.97634  630.99s\n",
      "    311       0.08240       0.07104      1.15989      0.97825  637.45s\n",
      "    312       0.08281       0.07122      1.16275      0.97809  631.34s\n",
      "    313       \u001b[36m0.08189\u001b[0m       0.07195      1.13822      0.97864  630.54s\n",
      "    314       0.08314       0.07030      1.18277      0.97915  630.85s\n",
      "    315       0.08248       \u001b[32m0.06917\u001b[0m      1.19238      0.97898  630.84s\n",
      "    316       0.08195       0.07490      1.09419      0.97735  631.65s\n",
      "    317       0.08223       0.07162      1.14805      0.97806  631.75s\n",
      "    318       0.08205       0.07539      1.08831      0.97761  627.13s\n",
      "    319       0.08193       0.07369      1.11187      0.97799  625.19s\n",
      "    320       0.08288       0.07078      1.17097      0.97867  624.99s\n",
      "    321       \u001b[36m0.08063\u001b[0m       0.07005      1.15117      0.97961  623.08s\n",
      "    322       0.08146       0.07087      1.14938      0.97855  621.26s\n",
      "    323       0.08150       0.07105      1.14714      0.97892  618.74s\n",
      "    324       0.08113       0.07388      1.09811      0.97761  618.86s\n",
      "    325       0.08178       0.07300      1.12029      0.97831  618.69s\n",
      "    326       0.08085       0.07541      1.07216      0.97660  618.80s\n",
      "    327       \u001b[36m0.08048\u001b[0m       0.07023      1.14603      0.97914  618.72s\n",
      "    328       0.08068       \u001b[32m0.06913\u001b[0m      1.16707      0.97875  618.69s\n",
      "    329       0.08139       0.07500      1.08522      0.97738  618.08s\n",
      "    330       \u001b[36m0.08022\u001b[0m       0.07446      1.07742      0.97728  618.02s\n",
      "    331       0.08071       0.07443      1.08438      0.97716  617.88s\n",
      "    332       0.08060       0.07198      1.11980      0.97857  617.81s\n",
      "    333       0.08053       0.07239      1.11237      0.97821  617.63s\n",
      "    334       0.08078       0.07236      1.11629      0.97863  617.92s\n",
      "    335       \u001b[36m0.07978\u001b[0m       0.07612      1.04798      0.97704  617.85s\n",
      "    336       \u001b[36m0.07949\u001b[0m       0.07183      1.10670      0.97824  617.96s\n",
      "    337       0.08037       0.07108      1.13073      0.97864  617.67s\n",
      "    338       0.08162       \u001b[32m0.06799\u001b[0m      1.20046      0.97972  617.93s\n",
      "    339       0.08005       0.07268      1.10137      0.97813  618.27s\n",
      "    340       0.08006       0.07497      1.06799      0.97761  618.26s\n",
      "    341       \u001b[36m0.07874\u001b[0m       0.07399      1.06412      0.97853  618.24s\n",
      "    342       0.08035       0.07282      1.10330      0.97796  618.33s\n",
      "    343       0.07988       0.07334      1.08919      0.97791  618.23s\n",
      "    344       0.07977       0.07146      1.11637      0.97853  618.33s\n",
      "    345       \u001b[36m0.07843\u001b[0m       0.07031      1.11543      0.97897  618.19s\n",
      "    346       0.07872       0.07091      1.11011      0.97854  618.28s\n",
      "    347       0.07992       0.06812      1.17316      0.97976  618.19s\n",
      "    348       0.07874       0.07076      1.11268      0.97955  618.22s\n",
      "    349       0.08013       0.07039      1.13840      0.97864  618.34s\n",
      "    350       0.07913       0.07159      1.10530      0.97842  618.20s\n",
      "    351       0.07924       0.07088      1.11793      0.97876  618.18s\n",
      "    352       0.07918       0.07186      1.10187      0.97815  618.21s\n",
      "    353       0.07890       0.06841      1.15333      0.98024  618.17s\n",
      "    354       0.07863       \u001b[32m0.06667\u001b[0m      1.17933      0.98060  618.28s\n",
      "    355       0.07882       0.06944      1.13508      0.97907  618.03s\n",
      "    356       0.07961       0.07224      1.10206      0.97854  617.93s\n",
      "    357       0.07901       0.07166      1.10256      0.97816  618.02s\n",
      "    358       0.07855       0.06975      1.12611      0.97908  617.99s\n",
      "    359       0.07868       0.06929      1.13553      0.97893  618.01s\n",
      "    360       \u001b[36m0.07832\u001b[0m       0.07361      1.06401      0.97832  617.92s\n",
      "    361       0.07859       0.06884      1.14149      0.98034  618.01s\n",
      "    362       \u001b[36m0.07768\u001b[0m       0.06952      1.11735      0.97980  618.00s\n",
      "    363       0.07858       0.06992      1.12383      0.97984  617.96s\n",
      "    364       \u001b[36m0.07767\u001b[0m       0.06955      1.11671      0.97965  618.10s\n",
      "    365       0.07770       0.07016      1.10752      0.97972  618.01s\n",
      "    366       \u001b[36m0.07696\u001b[0m       0.07156      1.07552      0.97875  618.13s\n",
      "    367       0.07815       0.07078      1.10413      0.97868  617.96s\n",
      "    368       0.07740       0.07290      1.06173      0.97845  618.13s\n",
      "    369       0.07748       0.07297      1.06192      0.97911  618.05s\n",
      "    370       0.07759       0.07150      1.08518      0.97940  618.10s\n",
      "    371       0.07743       0.06973      1.11051      0.98004  618.06s\n",
      "    372       0.07769       0.06677      1.16347      0.98070  618.11s\n",
      "    373       \u001b[36m0.07690\u001b[0m       0.07054      1.09016      0.97967  618.34s\n",
      "    374       0.07766       0.06899      1.12570      0.98042  618.14s\n",
      "    375       \u001b[36m0.07677\u001b[0m       0.07051      1.08882      0.97925  618.27s\n",
      "    376       0.07694       0.07136      1.07823      0.97927  618.30s\n",
      "    377       0.07680       0.07117      1.07910      0.97932  618.19s\n",
      "    378       0.07813       0.06902      1.13193      0.98042  618.17s\n",
      "    379       0.07801       0.07025      1.11049      0.98016  618.18s\n",
      "    380       0.07826       0.06972      1.12251      0.97983  618.26s\n",
      "    381       0.07706       0.07202      1.06987      0.97892  618.17s\n",
      "    382       \u001b[36m0.07661\u001b[0m       0.07109      1.07766      0.97951  618.14s\n",
      "    383       0.07710       0.06761      1.14031      0.98095  618.24s\n",
      "    384       0.07726       \u001b[32m0.06659\u001b[0m      1.16026      0.98048  618.18s\n",
      "    385       0.07675       0.07074      1.08499      0.97903  618.38s\n",
      "    386       0.07705       0.06915      1.11416      0.97980  618.47s\n",
      "    387       \u001b[36m0.07658\u001b[0m       0.07229      1.05927      0.97945  618.43s\n",
      "    388       0.07740       0.06996      1.10640      0.97978  620.73s\n",
      "    389       \u001b[36m0.07583\u001b[0m       0.07099      1.06822      0.97963  627.17s\n",
      "    390       0.07645       0.06979      1.09544      0.97989  631.88s\n",
      "    391       0.07733       0.07026      1.10058      0.97985  638.99s\n",
      "    392       0.07642       0.07100      1.07634      0.97908  639.38s\n",
      "    393       0.07661       0.07179      1.06708      0.97910  638.01s\n",
      "    394       0.07669       0.06776      1.13178      0.98010  639.01s\n",
      "    395       0.07604       0.07002      1.08601      0.97955  639.37s\n",
      "    396       0.07647       0.06886      1.11054      0.98026  640.86s\n",
      "    397       0.07653       0.07126      1.07390      0.97934  642.01s\n",
      "    398       \u001b[36m0.07556\u001b[0m       0.07094      1.06510      0.97948  642.28s\n",
      "    399       0.07557       0.07027      1.07551      0.97981  647.14s\n",
      "    400       \u001b[36m0.07534\u001b[0m       0.06987      1.07825      0.98007  640.93s\n",
      "    401       0.07610       0.07223      1.05355      0.97976  640.31s\n",
      "    402       0.07575       0.06806      1.11292      0.98088  640.41s\n",
      "    403       0.07572       0.07110      1.06496      0.97983  643.93s\n",
      "    404       0.07546       0.06976      1.08176      0.97995  639.27s\n",
      "    405       \u001b[36m0.07433\u001b[0m       0.07154      1.03891      0.97939  639.29s\n",
      "    406       0.07533       0.06697      1.12483      0.98096  638.75s\n",
      "    407       0.07596       0.06869      1.10593      0.98089  636.08s\n",
      "    408       0.07500       0.06894      1.08790      0.98056  638.09s\n",
      "    409       0.07474       0.06850      1.09107      0.98063  632.74s\n",
      "    410       0.07482       0.06975      1.07264      0.98018  636.66s\n",
      "    411       0.07503       0.07128      1.05273      0.97998  633.08s\n",
      "    412       0.07584       0.07192      1.05456      0.97973  639.10s\n",
      "    413       \u001b[36m0.07408\u001b[0m       0.06715      1.10318      0.98123  632.71s\n",
      "    414       0.07443       0.07058      1.05450      0.97963  635.04s\n",
      "Early stopping.\n",
      "Best valid loss was 0.066586 at epoch 384.\n",
      "Loaded parameters to layer 'image_conv1' (shape 24x1x13x13).\n",
      "Loaded parameters to layer 'image_conv1' (shape 24).\n",
      "Loaded parameters to layer 'image_conv2' (shape 22x24x13x13).\n",
      "Loaded parameters to layer 'image_conv2' (shape 22).\n",
      "Loaded parameters to layer 'prob_conv1' (shape 24x1x13x13).\n",
      "Loaded parameters to layer 'prob_conv1' (shape 24).\n",
      "Loaded parameters to layer 'prob_conv2' (shape 22x24x13x13).\n",
      "Loaded parameters to layer 'prob_conv2' (shape 22).\n",
      "Loaded parameters to layer 'binary_conv1' (shape 24x1x13x13).\n",
      "Loaded parameters to layer 'binary_conv1' (shape 24).\n",
      "Loaded parameters to layer 'binary_conv2' (shape 22x24x13x13).\n",
      "Loaded parameters to layer 'binary_conv2' (shape 22).\n",
      "Loaded parameters to layer 'hidden3' (shape 5346x256).\n",
      "Loaded parameters to layer 'hidden3' (shape 256).\n",
      "Loaded parameters to layer 'output' (shape 256x2).\n",
      "Loaded parameters to layer 'output' (shape 2).\n"
     ]
    }
   ],
   "source": [
    "cnn = cnn.fit({'image_input': X_train['image'],\n",
    "               'prob_input': X_train['prob'],\n",
    "               'binary_input': X_train['merged_array']},\n",
    "               y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9023717183770883"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# this is the score on cylinder_small1 with regularization and patience counter 50 and data augmentation\n",
    "cnn.score({'image_input': X_test['image'],\n",
    "           'prob_input': X_test['prob'],\n",
    "           'binary_input': X_test['merged_array']},\n",
    "           y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'matplotlib.pyplot' from '/home/d/nolearn/local/lib/python2.7/site-packages/matplotlib/pyplot.pyc'>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAEPCAYAAABsj5JaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7d3d5f6dd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this is the plot on cylinder_small1 with regularization and patience counter 50 and data augmentation\n",
    "plot_loss(cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import classification_report, accuracy_score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_true, y_pred = y_test, cnn.predict({'image_input': X_test['image'],\n",
    "           'prob_input': X_test['prob'],\n",
    "           'binary_input': X_test['merged_array']})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.92      0.89      0.90     13408\n",
      "          1       0.89      0.92      0.91     13408\n",
      "\n",
      "avg / total       0.91      0.91      0.91     26816\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_true, y_pred))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "with open('/home/d/nets/MergeNet_cylinder1_nooverlap_usevars/net.p', 'wb') as f:\n",
    "    pickle.dump(cnn, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import cPickle as pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sys.setrecursionlimit(10000000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}