MergeNet_image_prob_mergedarray_borderoverlap.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, border_overlap.\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "cnn = nets.MergeNet()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded /home/d/patches/cylinder_small1/ in 0.000457048416138 seconds.\n"
     ]
    }
   ],
   "source": [
    "X_train, y_train, X_test, y_test = mlp.Patch.load('cylinder_small1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# Neural Network with 1511874 learnable parameters\n",
      "\n",
      "## Layer information\n",
      "\n",
      "  #  name          size\n",
      "---  ------------  --------\n",
      "  0  image_input   1x75x75\n",
      "  1  image_conv1   16x63x63\n",
      "  2  image_pool1   16x31x31\n",
      "  3  image_conv2   16x19x19\n",
      "  4  image_pool2   16x9x9\n",
      "  5  prob_input    1x75x75\n",
      "  6  prob_conv1    16x63x63\n",
      "  7  prob_pool1    16x31x31\n",
      "  8  prob_conv2    16x19x19\n",
      "  9  prob_pool2    16x9x9\n",
      " 10  binary_input  1x75x75\n",
      " 11  binary_conv1  16x63x63\n",
      " 12  binary_pool1  16x31x31\n",
      " 13  binary_conv2  16x19x19\n",
      " 14  binary_pool2  16x9x9\n",
      " 15  border_input  1x75x75\n",
      " 16  border_conv1  16x63x63\n",
      " 17  border_pool1  16x31x31\n",
      " 18  border_conv2  16x19x19\n",
      " 19  border_pool2  16x9x9\n",
      " 20  merge         64x9x9\n",
      " 21  hidden3       256\n",
      " 22  dropout3      256\n",
      " 23  output        2\n",
      "\n",
      "  epoch    train loss    valid loss    train/val    valid acc  dur\n",
      "-------  ------------  ------------  -----------  -----------  ------\n",
      "      1       \u001b[36m0.58265\u001b[0m       \u001b[32m0.48079\u001b[0m      1.21186      0.76618  29.77s\n",
      "      2       \u001b[36m0.47341\u001b[0m       \u001b[32m0.42784\u001b[0m      1.10649      0.80279  29.71s\n",
      "      3       \u001b[36m0.44270\u001b[0m       \u001b[32m0.40949\u001b[0m      1.08109      0.81868  29.69s\n",
      "      4       \u001b[36m0.42611\u001b[0m       \u001b[32m0.39956\u001b[0m      1.06643      0.82164  29.74s\n",
      "      5       \u001b[36m0.41491\u001b[0m       \u001b[32m0.39415\u001b[0m      1.05266      0.82624  29.71s\n",
      "      6       \u001b[36m0.40692\u001b[0m       \u001b[32m0.38562\u001b[0m      1.05525      0.82853  29.96s\n",
      "      7       \u001b[36m0.40075\u001b[0m       \u001b[32m0.37833\u001b[0m      1.05925      0.83171  30.03s\n",
      "      8       \u001b[36m0.39111\u001b[0m       \u001b[32m0.37608\u001b[0m      1.03996      0.83641  30.04s\n",
      "      9       \u001b[36m0.38511\u001b[0m       \u001b[32m0.36870\u001b[0m      1.04451      0.84220  30.11s\n",
      "     10       \u001b[36m0.38129\u001b[0m       \u001b[32m0.36856\u001b[0m      1.03454      0.84083  30.06s\n",
      "     11       \u001b[36m0.37300\u001b[0m       \u001b[32m0.36150\u001b[0m      1.03179      0.84351  30.00s\n",
      "     12       \u001b[36m0.37026\u001b[0m       \u001b[32m0.35894\u001b[0m      1.03154      0.84521  30.01s\n",
      "     13       \u001b[36m0.36334\u001b[0m       \u001b[32m0.34814\u001b[0m      1.04367      0.85242  30.01s\n",
      "     14       \u001b[36m0.36238\u001b[0m       \u001b[32m0.34775\u001b[0m      1.04206      0.85341  30.01s\n",
      "     15       \u001b[36m0.35943\u001b[0m       \u001b[32m0.34575\u001b[0m      1.03958      0.85231  30.03s\n",
      "     16       \u001b[36m0.35488\u001b[0m       \u001b[32m0.33784\u001b[0m      1.05046      0.85494  30.00s\n",
      "     17       \u001b[36m0.34984\u001b[0m       0.33922      1.03133      0.85581  30.07s\n",
      "     18       \u001b[36m0.34785\u001b[0m       \u001b[32m0.33326\u001b[0m      1.04379      0.86068  30.34s\n",
      "     19       \u001b[36m0.34303\u001b[0m       \u001b[32m0.32967\u001b[0m      1.04054      0.86352  30.31s\n",
      "     20       \u001b[36m0.34045\u001b[0m       0.34245      0.99415      0.85215  30.33s\n",
      "     21       \u001b[36m0.33977\u001b[0m       \u001b[32m0.32900\u001b[0m      1.03273      0.86428  30.35s\n",
      "     22       \u001b[36m0.33517\u001b[0m       \u001b[32m0.32702\u001b[0m      1.02491      0.86187  30.33s\n",
      "     23       \u001b[36m0.33368\u001b[0m       \u001b[32m0.31908\u001b[0m      1.04577      0.86843  30.41s\n",
      "     24       \u001b[36m0.33154\u001b[0m       0.32138      1.03163      0.86532  30.34s\n",
      "     25       \u001b[36m0.32840\u001b[0m       \u001b[32m0.31838\u001b[0m      1.03148      0.86778  30.39s\n",
      "     26       0.33070       0.33070      0.99999      0.85942  30.35s\n",
      "     27       \u001b[36m0.32314\u001b[0m       0.32423      0.99663      0.86166  30.35s\n",
      "     28       0.32381       0.31909      1.01480      0.86560  30.34s\n",
      "     29       \u001b[36m0.32055\u001b[0m       \u001b[32m0.31459\u001b[0m      1.01897      0.86975  30.35s\n",
      "     30       \u001b[36m0.31647\u001b[0m       \u001b[32m0.30080\u001b[0m      1.05210      0.87275  30.35s\n",
      "     31       \u001b[36m0.31437\u001b[0m       0.30088      1.04481      0.87232  30.35s\n",
      "     32       0.31581       0.30558      1.03346      0.87226  30.36s\n",
      "     33       \u001b[36m0.31053\u001b[0m       0.30600      1.01478      0.87205  30.34s\n",
      "     34       \u001b[36m0.30944\u001b[0m       0.30319      1.02062      0.87527  30.34s\n",
      "     35       \u001b[36m0.30630\u001b[0m       0.30353      1.00913      0.87418  30.34s\n",
      "     36       \u001b[36m0.30389\u001b[0m       0.30639      0.99185      0.86937  30.37s\n",
      "     37       0.30688       \u001b[32m0.30024\u001b[0m      1.02209      0.87669  30.37s\n",
      "     38       \u001b[36m0.30027\u001b[0m       \u001b[32m0.29596\u001b[0m      1.01457      0.87538  30.39s\n",
      "     39       \u001b[36m0.29731\u001b[0m       \u001b[32m0.29467\u001b[0m      1.00896      0.87784  30.35s\n",
      "     40       0.29975       \u001b[32m0.29039\u001b[0m      1.03224      0.87500  30.37s\n",
      "     41       \u001b[36m0.29367\u001b[0m       0.29156      1.00723      0.87844  30.36s\n",
      "     42       \u001b[36m0.29135\u001b[0m       \u001b[32m0.28624\u001b[0m      1.01784      0.88003  30.40s\n",
      "     43       0.29137       \u001b[32m0.28278\u001b[0m      1.03035      0.88451  30.36s\n",
      "     44       \u001b[36m0.28719\u001b[0m       0.28859      0.99515      0.88079  30.40s\n",
      "     45       \u001b[36m0.28568\u001b[0m       \u001b[32m0.27874\u001b[0m      1.02490      0.88445  30.35s\n",
      "     46       \u001b[36m0.28471\u001b[0m       0.28570      0.99652      0.88057  30.39s\n",
      "     47       \u001b[36m0.28171\u001b[0m       \u001b[32m0.27867\u001b[0m      1.01094      0.88522  30.37s\n",
      "     48       \u001b[36m0.28048\u001b[0m       0.28262      0.99242      0.88205  30.38s\n",
      "     49       \u001b[36m0.27922\u001b[0m       0.28188      0.99057      0.88112  30.37s\n",
      "     50       \u001b[36m0.27561\u001b[0m       \u001b[32m0.27319\u001b[0m      1.00887      0.88620  30.36s\n",
      "     51       0.27682       0.28333      0.97700      0.87642  30.38s\n",
      "     52       \u001b[36m0.26995\u001b[0m       \u001b[32m0.27296\u001b[0m      0.98897      0.88615  30.39s\n",
      "     53       0.27292       \u001b[32m0.27001\u001b[0m      1.01078      0.88483  30.35s\n",
      "     54       0.27022       0.27138      0.99574      0.88468  30.37s\n",
      "     55       \u001b[36m0.26817\u001b[0m       \u001b[32m0.26333\u001b[0m      1.01840      0.89195  30.38s\n",
      "     56       \u001b[36m0.26423\u001b[0m       0.26856      0.98388      0.88495  30.38s\n",
      "     57       \u001b[36m0.26252\u001b[0m       0.27046      0.97064      0.88702  30.38s\n",
      "     58       \u001b[36m0.26005\u001b[0m       \u001b[32m0.26032\u001b[0m      0.99897      0.89030  30.39s\n",
      "     59       0.26032       \u001b[32m0.25973\u001b[0m      1.00226      0.89156  30.36s\n",
      "     60       \u001b[36m0.25688\u001b[0m       0.26354      0.97475      0.88997  30.36s\n",
      "     61       \u001b[36m0.25563\u001b[0m       0.26800      0.95384      0.88714  30.53s\n",
      "     62       \u001b[36m0.25207\u001b[0m       \u001b[32m0.25363\u001b[0m      0.99386      0.89260  30.70s\n",
      "     63       \u001b[36m0.25178\u001b[0m       \u001b[32m0.25156\u001b[0m      1.00086      0.89588  30.70s\n",
      "     64       \u001b[36m0.24999\u001b[0m       0.25711      0.97231      0.89206  30.73s\n",
      "     65       \u001b[36m0.24908\u001b[0m       \u001b[32m0.24697\u001b[0m      1.00853      0.89872  30.73s\n",
      "     66       \u001b[36m0.24516\u001b[0m       0.24960      0.98221      0.89681  30.73s\n",
      "     67       0.24637       0.24919      0.98866      0.89538  30.71s\n",
      "     68       \u001b[36m0.23939\u001b[0m       \u001b[32m0.24123\u001b[0m      0.99238      0.90254  30.70s\n",
      "     69       0.23981       \u001b[32m0.23942\u001b[0m      1.00161      0.90402  30.69s\n",
      "     70       \u001b[36m0.23827\u001b[0m       \u001b[32m0.23693\u001b[0m      1.00565      0.90145  30.71s\n",
      "     71       \u001b[36m0.23473\u001b[0m       \u001b[32m0.23648\u001b[0m      0.99259      0.90506  30.73s\n",
      "     72       \u001b[36m0.23351\u001b[0m       0.24570      0.95037      0.90063  30.71s\n",
      "     73       \u001b[36m0.23265\u001b[0m       0.23711      0.98120      0.89856  30.73s\n",
      "     74       \u001b[36m0.23235\u001b[0m       0.23977      0.96902      0.90161  30.72s\n",
      "     75       \u001b[36m0.23217\u001b[0m       0.23704      0.97943      0.90413  30.71s\n",
      "     76       \u001b[36m0.22710\u001b[0m       \u001b[32m0.22571\u001b[0m      1.00614      0.91063  30.72s\n",
      "     77       \u001b[36m0.22446\u001b[0m       0.23222      0.96658      0.90495  30.70s\n",
      "     78       \u001b[36m0.21943\u001b[0m       0.23387      0.93825      0.90451  30.68s\n",
      "     79       \u001b[36m0.21691\u001b[0m       \u001b[32m0.22093\u001b[0m      0.98180      0.91156  30.68s\n",
      "     80       0.22013       0.22382      0.98351      0.90872  30.73s\n",
      "     81       \u001b[36m0.21458\u001b[0m       0.22160      0.96833      0.91085  30.71s\n",
      "     82       \u001b[36m0.21243\u001b[0m       0.22897      0.92777      0.90583  30.70s\n",
      "     83       0.21383       0.22735      0.94055      0.90981  30.70s\n",
      "     84       \u001b[36m0.20695\u001b[0m       \u001b[32m0.21966\u001b[0m      0.94214      0.90938  30.71s\n",
      "     85       0.20832       \u001b[32m0.21515\u001b[0m      0.96827      0.91370  30.69s\n",
      "     86       \u001b[36m0.20691\u001b[0m       0.21972      0.94173      0.91216  30.69s\n",
      "     87       \u001b[36m0.19865\u001b[0m       \u001b[32m0.20868\u001b[0m      0.95191      0.91462  30.70s\n",
      "     88       0.19921       0.22148      0.89948      0.91052  30.73s\n",
      "     89       0.19966       0.21286      0.93798      0.91353  30.71s\n",
      "     90       \u001b[36m0.19681\u001b[0m       \u001b[32m0.20181\u001b[0m      0.97524      0.92299  30.72s\n",
      "     91       \u001b[36m0.19171\u001b[0m       0.20267      0.94589      0.92243  30.71s\n",
      "     92       0.19392       \u001b[32m0.19631\u001b[0m      0.98779      0.92244  30.71s\n",
      "     93       \u001b[36m0.19027\u001b[0m       0.20369      0.93414      0.91883  30.71s\n",
      "     94       \u001b[36m0.18697\u001b[0m       0.20582      0.90841      0.91960  30.70s\n",
      "     95       \u001b[36m0.18524\u001b[0m       \u001b[32m0.19555\u001b[0m      0.94730      0.92375  30.71s\n",
      "     96       \u001b[36m0.18270\u001b[0m       0.20653      0.88462      0.91998  30.70s\n",
      "     97       \u001b[36m0.17995\u001b[0m       \u001b[32m0.19287\u001b[0m      0.93304      0.92156  30.69s\n",
      "     98       \u001b[36m0.17762\u001b[0m       \u001b[32m0.18760\u001b[0m      0.94680      0.92834  30.73s\n",
      "     99       \u001b[36m0.17692\u001b[0m       0.19293      0.91704      0.92626  30.70s\n",
      "    100       \u001b[36m0.17361\u001b[0m       0.19478      0.89132      0.92332  30.70s\n",
      "    101       0.17853       0.18782      0.95056      0.92588  30.70s\n",
      "    102       0.17535       \u001b[32m0.18472\u001b[0m      0.94925      0.93075  30.70s\n",
      "    103       \u001b[36m0.17053\u001b[0m       \u001b[32m0.18276\u001b[0m      0.93309      0.93141  30.72s\n",
      "    104       \u001b[36m0.16839\u001b[0m       0.18653      0.90276      0.92851  30.72s\n",
      "    105       \u001b[36m0.16324\u001b[0m       \u001b[32m0.18035\u001b[0m      0.90510      0.93233  30.69s\n",
      "    106       \u001b[36m0.16246\u001b[0m       \u001b[32m0.16920\u001b[0m      0.96018      0.93217  30.70s\n",
      "    107       0.16327       0.17588      0.92831      0.93315  30.70s\n",
      "    108       \u001b[36m0.15929\u001b[0m       0.16977      0.93823      0.93682  30.69s\n",
      "    109       \u001b[36m0.15706\u001b[0m       0.17431      0.90100      0.93183  30.70s\n",
      "    110       \u001b[36m0.15570\u001b[0m       \u001b[32m0.16591\u001b[0m      0.93844      0.93856  30.69s\n",
      "    111       \u001b[36m0.15105\u001b[0m       \u001b[32m0.16483\u001b[0m      0.91637      0.93970  30.74s\n",
      "    112       \u001b[36m0.14601\u001b[0m       \u001b[32m0.16122\u001b[0m      0.90565      0.94179  30.72s\n",
      "    113       0.14927       0.16392      0.91065      0.93873  30.68s\n",
      "    114       0.14797       0.16246      0.91082      0.93992  30.69s\n",
      "    115       0.14607       \u001b[32m0.16001\u001b[0m      0.91292      0.94299  30.69s\n",
      "    116       \u001b[36m0.14082\u001b[0m       0.16158      0.87151      0.93992  30.88s\n",
      "    117       0.14162       \u001b[32m0.15659\u001b[0m      0.90442      0.94294  30.81s\n",
      "    118       0.14118       \u001b[32m0.14462\u001b[0m      0.97618      0.94829  33.33s\n",
      "    119       \u001b[36m0.13543\u001b[0m       0.14816      0.91412      0.94550  30.55s\n",
      "    120       0.13832       0.15007      0.92172      0.94326  30.77s\n",
      "    121       \u001b[36m0.13399\u001b[0m       0.14923      0.89785      0.94452  30.47s\n",
      "    122       \u001b[36m0.13391\u001b[0m       0.14473      0.92525      0.94944  30.34s\n",
      "    123       \u001b[36m0.13360\u001b[0m       \u001b[32m0.14269\u001b[0m      0.93632      0.95065  30.49s\n",
      "    124       \u001b[36m0.13274\u001b[0m       \u001b[32m0.14070\u001b[0m      0.94339      0.94846  30.46s\n",
      "    125       \u001b[36m0.12403\u001b[0m       \u001b[32m0.13562\u001b[0m      0.91453      0.95375  30.39s\n",
      "    126       0.12733       0.13854      0.91906      0.95397  30.47s\n",
      "    127       0.12554       0.13977      0.89820      0.95195  30.74s\n",
      "    128       0.12729       0.13732      0.92696      0.95004  30.81s\n",
      "    129       0.12504       0.14453      0.86515      0.94971  31.01s\n",
      "    130       \u001b[36m0.11946\u001b[0m       \u001b[32m0.13053\u001b[0m      0.91520      0.95807  30.67s\n",
      "    131       \u001b[36m0.11637\u001b[0m       \u001b[32m0.12845\u001b[0m      0.90597      0.95764  30.68s\n",
      "    132       \u001b[36m0.11487\u001b[0m       \u001b[32m0.12762\u001b[0m      0.90007      0.95780  30.68s\n",
      "    133       \u001b[36m0.11150\u001b[0m       \u001b[32m0.12473\u001b[0m      0.89396      0.95725  30.62s\n",
      "    134       \u001b[36m0.10857\u001b[0m       0.13972      0.77709      0.95004  30.65s\n",
      "    135       0.11218       0.13315      0.84255      0.95381  30.67s\n",
      "    136       0.11533       0.12770      0.90317      0.95496  30.71s\n",
      "    137       0.11080       0.13044      0.84943      0.95047  30.67s\n",
      "    138       0.11093       \u001b[32m0.12377\u001b[0m      0.89627      0.96081  30.69s\n",
      "    139       \u001b[36m0.10475\u001b[0m       0.12475      0.83968      0.95944  30.72s\n",
      "    140       \u001b[36m0.10206\u001b[0m       \u001b[32m0.11741\u001b[0m      0.86922      0.96266  30.69s\n",
      "    141       0.10524       \u001b[32m0.11424\u001b[0m      0.92116      0.96343  30.06s\n",
      "    142       \u001b[36m0.09977\u001b[0m       \u001b[32m0.11307\u001b[0m      0.88230      0.96409  30.71s\n",
      "    143       0.10236       0.11531      0.88765      0.96174  30.82s\n",
      "    144       \u001b[36m0.09510\u001b[0m       \u001b[32m0.11290\u001b[0m      0.84228      0.96479  31.08s\n",
      "    145       0.09939       \u001b[32m0.11140\u001b[0m      0.89215      0.96354  30.69s\n",
      "    146       0.09739       \u001b[32m0.10970\u001b[0m      0.88782      0.96688  31.11s\n",
      "    147       \u001b[36m0.09454\u001b[0m       0.11297      0.83679      0.96518  30.74s\n",
      "    148       \u001b[36m0.09318\u001b[0m       \u001b[32m0.10859\u001b[0m      0.85809      0.96502  30.64s\n",
      "    149       0.09500       \u001b[32m0.10431\u001b[0m      0.91077      0.96928  30.72s\n",
      "    150       \u001b[36m0.08417\u001b[0m       0.10800      0.77932      0.96479  30.79s\n",
      "    151       0.08788       0.11134      0.78933      0.96381  31.14s\n",
      "    152       0.08808       0.10645      0.82742      0.96791  30.67s\n",
      "    153       0.08455       \u001b[32m0.10112\u001b[0m      0.83611      0.97043  30.65s\n",
      "    154       \u001b[36m0.08390\u001b[0m       0.10347      0.81085      0.96907  30.68s\n",
      "    155       \u001b[36m0.08372\u001b[0m       \u001b[32m0.09859\u001b[0m      0.84912      0.97010  30.65s\n",
      "    156       0.08856       \u001b[32m0.09722\u001b[0m      0.91095      0.97245  30.67s\n",
      "    157       0.08552       0.10032      0.85253      0.97081  30.72s\n",
      "    158       \u001b[36m0.08339\u001b[0m       0.09949      0.83815      0.97065  30.89s\n",
      "    159       \u001b[36m0.07634\u001b[0m       \u001b[32m0.09424\u001b[0m      0.81009      0.97425  30.67s\n",
      "    160       0.08235       \u001b[32m0.09356\u001b[0m      0.88015      0.97185  30.81s\n",
      "    161       0.07998       0.09966      0.80254      0.96934  30.73s\n",
      "    162       0.07721       \u001b[32m0.09030\u001b[0m      0.85505      0.97524  30.64s\n",
      "    163       0.07743       0.09526      0.81278      0.97201  30.81s\n",
      "    164       \u001b[36m0.07595\u001b[0m       0.09173      0.82795      0.97420  30.66s\n",
      "    165       0.07731       0.09177      0.84237      0.97574  30.79s\n",
      "    166       \u001b[36m0.07151\u001b[0m       0.09390      0.76151      0.97343  30.84s\n",
      "    167       0.07624       \u001b[32m0.08845\u001b[0m      0.86196      0.97524  30.88s\n",
      "    168       0.07183       0.09345      0.76863      0.97305  30.67s\n",
      "    169       \u001b[36m0.06987\u001b[0m       0.08962      0.77959      0.97507  30.79s\n",
      "    170       0.07135       \u001b[32m0.08691\u001b[0m      0.82095      0.97600  30.72s\n",
      "    171       0.07064       \u001b[32m0.08572\u001b[0m      0.82410      0.97529  30.85s\n",
      "    172       \u001b[36m0.06888\u001b[0m       0.08777      0.78482      0.97387  30.73s\n",
      "    173       0.07242       \u001b[32m0.08386\u001b[0m      0.86354      0.97911  30.67s\n",
      "    174       \u001b[36m0.06652\u001b[0m       0.08644      0.76953      0.97829  30.83s\n",
      "    175       \u001b[36m0.06619\u001b[0m       \u001b[32m0.07974\u001b[0m      0.83010      0.97798  30.87s\n",
      "    176       \u001b[36m0.06605\u001b[0m       0.08745      0.75528      0.97589  31.18s\n",
      "    177       \u001b[36m0.06540\u001b[0m       0.08272      0.79066      0.97808  30.62s\n",
      "    178       \u001b[36m0.06145\u001b[0m       \u001b[32m0.07793\u001b[0m      0.78861      0.97829  30.83s\n",
      "    179       0.06603       0.08177      0.80746      0.97786  30.79s\n",
      "    180       \u001b[36m0.06104\u001b[0m       \u001b[32m0.07506\u001b[0m      0.81315      0.97934  32.49s\n",
      "    181       0.06240       0.08186      0.76229      0.97747  186.23s\n",
      "    182       \u001b[36m0.06043\u001b[0m       0.07552      0.80030      0.98060  30.50s\n",
      "    183       0.06292       0.07791      0.80764      0.97944  31.05s\n",
      "    184       0.06254       0.08430      0.74187      0.98016  30.18s\n",
      "    185       0.06433       0.08038      0.80027      0.97665  30.17s\n",
      "    186       \u001b[36m0.05620\u001b[0m       0.07666      0.73301      0.97901  30.18s\n",
      "    187       0.05653       0.07891      0.71639      0.97780  30.15s\n",
      "    188       0.05659       0.08058      0.70229      0.97802  30.15s\n",
      "    189       0.05707       \u001b[32m0.07327\u001b[0m      0.77880      0.98185  30.30s\n",
      "    190       \u001b[36m0.05452\u001b[0m       0.08212      0.66389      0.97884  30.27s\n",
      "    191       \u001b[36m0.05439\u001b[0m       0.07513      0.72400      0.98081  30.57s\n",
      "    192       \u001b[36m0.05273\u001b[0m       0.07403      0.71230      0.98322  30.82s\n",
      "    193       \u001b[36m0.05219\u001b[0m       0.07787      0.67024      0.98272  30.78s\n",
      "    194       0.05331       \u001b[32m0.06787\u001b[0m      0.78548      0.98431  30.87s\n",
      "    195       \u001b[36m0.05099\u001b[0m       0.07278      0.70062      0.98229  30.48s\n",
      "    196       0.05133       0.07103      0.72261      0.98196  30.46s\n",
      "    197       0.05237       0.07314      0.71597      0.98409  30.49s\n",
      "    198       0.05368       0.07595      0.70676      0.98021  30.47s\n",
      "    199       \u001b[36m0.04882\u001b[0m       0.07098      0.68779      0.98179  30.45s\n",
      "    200       0.05075       0.07140      0.71085      0.98147  30.52s\n",
      "    201       0.05026       0.07572      0.66373      0.98114  30.94s\n",
      "    202       \u001b[36m0.04790\u001b[0m       0.07404      0.64696      0.98425  30.14s\n",
      "    203       \u001b[36m0.04709\u001b[0m       0.07125      0.66090      0.98393  30.33s\n",
      "    204       0.04764       0.07094      0.67158      0.98393  30.83s\n",
      "    205       \u001b[36m0.04522\u001b[0m       0.07677      0.58902      0.98097  30.81s\n",
      "    206       0.04537       0.07099      0.63907      0.98212  30.22s\n",
      "    207       \u001b[36m0.04437\u001b[0m       0.07229      0.61374      0.98420  34.51s\n",
      "    208       0.04487       \u001b[32m0.06695\u001b[0m      0.67022      0.98409  34.85s\n",
      "    209       0.04477       0.07136      0.62738      0.98267  32.46s\n",
      "    210       0.04649       0.06809      0.68270      0.98376  31.92s\n",
      "    211       0.04667       0.06892      0.67711      0.98458  33.21s\n",
      "    212       0.04463       0.06733      0.66275      0.98409  33.65s\n",
      "    213       0.04807       0.07328      0.65603      0.98306  30.14s\n",
      "    214       0.04519       0.07256      0.62274      0.98163  33.64s\n",
      "    215       \u001b[36m0.04022\u001b[0m       0.07509      0.53563      0.98393  34.97s\n",
      "    216       0.04324       0.06723      0.64319      0.98393  30.17s\n",
      "    217       0.04488       \u001b[32m0.06572\u001b[0m      0.68285      0.98409  30.85s\n",
      "    218       0.04306       \u001b[32m0.06357\u001b[0m      0.67733      0.98556  31.79s\n",
      "    219       0.04092       0.06593      0.62059      0.98524  36.84s\n",
      "    220       0.04055       0.07718      0.52542      0.98327  36.20s\n",
      "    221       \u001b[36m0.03809\u001b[0m       0.07491      0.50842      0.98278  35.07s\n",
      "    222       0.03952       0.06643      0.59488      0.98524  35.40s\n",
      "    223       \u001b[36m0.03803\u001b[0m       0.07120      0.53422      0.98442  35.52s\n",
      "    224       \u001b[36m0.03650\u001b[0m       0.07236      0.50439      0.98245  34.74s\n",
      "    225       0.04192       0.08135      0.51533      0.98163  35.11s\n",
      "    226       0.04146       0.06722      0.61680      0.98366  35.29s\n",
      "    227       0.03941       0.06668      0.59101      0.98370  30.01s\n",
      "    228       0.04114       0.06407      0.64203      0.98485  30.16s\n",
      "    229       0.03734       0.06681      0.55894      0.98556  30.34s\n",
      "    230       0.04014       0.07796      0.51492      0.98409  30.35s\n",
      "    231       0.03665       0.06800      0.53902      0.98556  30.20s\n",
      "    232       0.04003       0.06403      0.62518      0.98568  30.44s\n",
      "    233       \u001b[36m0.03573\u001b[0m       0.07144      0.50013      0.98245  30.44s\n",
      "    234       \u001b[36m0.03531\u001b[0m       0.06530      0.54077      0.98781  30.49s\n",
      "    235       0.03734       0.06750      0.55320      0.98573  30.45s\n",
      "    236       0.03755       0.06672      0.56279      0.98518  30.49s\n",
      "    237       \u001b[36m0.03498\u001b[0m       \u001b[32m0.05969\u001b[0m      0.58610      0.98704  30.47s\n",
      "    238       \u001b[36m0.03465\u001b[0m       0.06171      0.56159      0.98759  30.48s\n",
      "    239       \u001b[36m0.03419\u001b[0m       0.06835      0.50014      0.98452  30.57s\n",
      "    240       \u001b[36m0.03234\u001b[0m       0.06268      0.51591      0.98540  30.42s\n",
      "    241       \u001b[36m0.02924\u001b[0m       0.06230      0.46925      0.98737  30.53s\n",
      "    242       \u001b[36m0.02920\u001b[0m       0.06797      0.42961      0.98677  30.53s\n",
      "    243       0.03435       0.06368      0.53943      0.98770  30.55s\n",
      "    244       0.03294       0.06532      0.50423      0.98649  30.58s\n",
      "    245       0.02937       0.06700      0.43840      0.98458  30.82s\n",
      "    246       \u001b[36m0.02837\u001b[0m       0.06565      0.43221      0.98688  30.44s\n",
      "    247       0.03096       0.06469      0.47862      0.98655  30.43s\n",
      "    248       0.03312       0.06933      0.47774      0.98513  30.43s\n",
      "    249       0.03066       0.06044      0.50729      0.98814  30.55s\n",
      "    250       0.02880       0.06365      0.45256      0.98704  30.67s\n",
      "    251       0.03317       0.06358      0.52172      0.98770  30.45s\n",
      "    252       0.03204       \u001b[32m0.05927\u001b[0m      0.54056      0.98792  30.46s\n",
      "    253       0.03225       0.06178      0.52191      0.98644  30.72s\n",
      "    254       0.03148       0.07247      0.43432      0.98573  30.63s\n",
      "    255       0.02941       0.06949      0.42324      0.98666  30.63s\n",
      "    256       \u001b[36m0.02759\u001b[0m       0.06351      0.43435      0.98770  30.70s\n",
      "    257       0.02790       0.06286      0.44378      0.98677  30.71s\n",
      "    258       0.02993       0.06339      0.47216      0.98747  30.80s\n",
      "    259       0.02803       0.05960      0.47032      0.98988  30.81s\n",
      "    260       0.02975       0.06883      0.43217      0.98671  30.79s\n",
      "    261       0.02776       0.06273      0.44257      0.98791  30.83s\n",
      "    262       0.02870       0.06337      0.45284      0.98966  30.78s\n",
      "    263       0.03004       0.06227      0.48242      0.98857  30.71s\n",
      "    264       \u001b[36m0.02696\u001b[0m       0.07262      0.37126      0.98688  30.53s\n",
      "    265       0.02969       0.06238      0.47600      0.98688  30.62s\n",
      "    266       0.02723       \u001b[32m0.05804\u001b[0m      0.46922      0.98835  30.52s\n",
      "    267       \u001b[36m0.02487\u001b[0m       0.08306      0.29944      0.98425  30.87s\n",
      "    268       0.02902       0.06746      0.43018      0.98704  30.86s\n",
      "    269       0.02574       0.06836      0.37650      0.98693  30.85s\n",
      "    270       0.02648       0.06603      0.40108      0.98720  30.79s\n",
      "    271       0.02548       0.06395      0.39847      0.98764  30.82s\n",
      "    272       0.03016       0.06151      0.49029      0.98819  30.43s\n",
      "    273       0.02960       0.06168      0.47986      0.98917  30.75s\n",
      "    274       \u001b[36m0.02406\u001b[0m       0.06282      0.38306      0.98819  30.77s\n",
      "    275       0.02649       0.06115      0.43314      0.98622  30.42s\n",
      "    276       0.02723       0.06884      0.39553      0.98573  30.48s\n",
      "    277       0.02602       0.06743      0.38590      0.98759  30.75s\n",
      "    278       0.02459       0.06462      0.38050      0.98704  30.82s\n",
      "    279       \u001b[36m0.02331\u001b[0m       0.07005      0.33273      0.98606  30.83s\n",
      "    280       0.02551       0.06261      0.40754      0.98819  31.16s\n",
      "    281       0.02796       0.06202      0.45090      0.98737  31.16s\n",
      "    282       0.02587       0.06873      0.37644      0.98726  31.00s\n",
      "    283       0.02656       0.06224      0.42673      0.98720  30.94s\n",
      "    284       0.02543       0.06794      0.37425      0.98743  31.06s\n",
      "    285       0.02578       0.06545      0.39384      0.98622  31.06s\n",
      "    286       0.02365       0.06138      0.38529      0.98890  30.77s\n",
      "    287       0.02407       0.06834      0.35224      0.98638  30.80s\n",
      "    288       0.02585       0.06576      0.39312      0.98638  30.78s\n",
      "    289       \u001b[36m0.02287\u001b[0m       0.07103      0.32199      0.98846  30.87s\n",
      "    290       0.02545       0.06528      0.38986      0.98770  30.92s\n",
      "    291       0.02355       0.06286      0.37461      0.98792  31.18s\n",
      "    292       \u001b[36m0.02198\u001b[0m       0.06418      0.34251      0.98655  31.19s\n",
      "    293       0.02354       0.06692      0.35185      0.98884  31.10s\n",
      "    294       0.02350       0.06885      0.34130      0.98720  30.89s\n",
      "    295       0.02243       0.06374      0.35197      0.98852  31.15s\n",
      "    296       \u001b[36m0.02087\u001b[0m       0.06435      0.32423      0.98753  30.91s\n",
      "    297       0.02153       0.06884      0.31272      0.98710  30.84s\n",
      "    298       0.02535       0.06229      0.40703      0.98770  30.77s\n",
      "    299       \u001b[36m0.02071\u001b[0m       0.07180      0.28844      0.98743  30.80s\n",
      "    300       0.02409       0.06998      0.34422      0.98753  30.98s\n",
      "    301       0.02228       0.06184      0.36029      0.98950  31.21s\n",
      "    302       0.02166       0.06622      0.32709      0.98808  31.21s\n",
      "    303       0.02250       0.06748      0.33340      0.98802  31.20s\n",
      "    304       0.02081       0.06914      0.30095      0.98852  31.21s\n",
      "    305       \u001b[36m0.01794\u001b[0m       0.06622      0.27087      0.98825  31.14s\n",
      "    306       0.02049       0.07120      0.28774      0.98671  31.15s\n",
      "    307       0.02003       0.07117      0.28149      0.98808  30.96s\n",
      "    308       0.02141       0.06768      0.31630      0.98802  30.88s\n",
      "    309       0.02065       0.06948      0.29726      0.98753  31.03s\n",
      "    310       0.01862       0.06545      0.28445      0.98770  30.96s\n",
      "    311       0.02098       0.06942      0.30216      0.98830  30.92s\n",
      "    312       0.02140       0.05858      0.36540      0.99070  31.05s\n",
      "    313       0.02036       0.07405      0.27492      0.98835  31.18s\n",
      "    314       0.02087       0.05867      0.35574      0.98988  31.20s\n",
      "    315       0.02058       0.06792      0.30299      0.98884  31.02s\n",
      "    316       0.01851       0.08180      0.22626      0.98573  31.19s\n",
      "Early stopping.\n",
      "Best valid loss was 0.058038 at epoch 266.\n",
      "Loaded parameters to layer 'image_conv1' (shape 16x1x13x13).\n",
      "Loaded parameters to layer 'image_conv1' (shape 16).\n",
      "Loaded parameters to layer 'image_conv2' (shape 16x16x13x13).\n",
      "Loaded parameters to layer 'image_conv2' (shape 16).\n",
      "Loaded parameters to layer 'prob_conv1' (shape 16x1x13x13).\n",
      "Loaded parameters to layer 'prob_conv1' (shape 16).\n",
      "Loaded parameters to layer 'prob_conv2' (shape 16x16x13x13).\n",
      "Loaded parameters to layer 'prob_conv2' (shape 16).\n",
      "Loaded parameters to layer 'binary_conv1' (shape 16x1x13x13).\n",
      "Loaded parameters to layer 'binary_conv1' (shape 16).\n",
      "Loaded parameters to layer 'binary_conv2' (shape 16x16x13x13).\n",
      "Loaded parameters to layer 'binary_conv2' (shape 16).\n",
      "Loaded parameters to layer 'border_conv1' (shape 16x1x13x13).\n",
      "Loaded parameters to layer 'border_conv1' (shape 16).\n",
      "Loaded parameters to layer 'border_conv2' (shape 16x16x13x13).\n",
      "Loaded parameters to layer 'border_conv2' (shape 16).\n",
      "Loaded parameters to layer 'hidden3' (shape 5184x256).\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",
    "               'border_input': X_train['border_overlap']},\n",
    "               y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.852569018404908"
      ]
     },
     "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",
    "           'border_input': X_test['border_overlap']},\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": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAEPCAYAAAC3NDh4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f647c777bd0>"
      ]
     },
     "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": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8312883435582822"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# this is the score on cylinder_small1 with regularization and patience counter 50\n",
    "cnn.score({'image_input': X_test['image'],\n",
    "           'prob_input': X_test['prob'],\n",
    "           'binary_input': X_test['merged_array'],\n",
    "           'border_input': X_test['border_overlap']},\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": "iVBORw0KGgoAAAANSUhEUgAAAZoAAAEPCAYAAAB7rQKTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njZmZJeWgMTOzpBw0ZmaWlIPGzMyS+v8UZnQTkAH73QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f2f4c320d90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this is the plot on cylinder_small1 with regularization and patience counter 50\n",
    "plot_loss(cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8224693251533742"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# this is the score on cylinder_small1 with regularization\n",
    "cnn.score({'image_input': X_test['image'],\n",
    "           'prob_input': X_test['prob'],\n",
    "           'binary_input': X_test['merged_array'],\n",
    "           'border_input': X_test['border_overlap']},\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": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAEPCAYAAAC3NDh4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3XlYlOX6wPHvDeKCCGquoSJuuIX7kqaSpbmbpuaeWqYd\nzTZL/XVM61idFttPlktm5pbmkpZpaajlnrmLO4jgvrCZIszz++MdCRQUlWGY4f5c11xn5p1n3rkf\n6HD77GKMQSmllEqPh7MDUEoplXNpklBKKZUhTRJKKaUypElCKaVUhjRJKKWUypAmCaWUUhlyeJIQ\nkTYiEiYiB0RkVAZlQkTkLxHZLSK/OTompZRSmSOOXCchIh7AAeAhIBrYAvQ0xoSlKuMHrAdaG2Oi\nRKSYMeasw4JSSimVaY5uSTQEDhpjIowxV4G5QOfryvQGvjfGRAFoglBKqZzD0UnCH4hM9fq4/Vpq\nVYCiIvKbiGwRkX4OjkkppVQm5XF2AFgx1AVaAgWBDSKywRhzyLlhKaWUcnSSiALKpXpdxn4ttePA\nWWPMZeCyiKwFagFpkoSI6CZTSil1B4wxcqefdXR30xagkogEiEheoCfww3VllgAPiIiniHgDjYB9\n6d3MGOO2j3Hjxjk9Bq2f1i+31S031O9uObQlYYxJFpHhwEqshDTNGLNPRIZYb5vJxpgwEVkB7ASS\ngcnGmL2OjEsppVTmOHxMwhjzMxB03bUvr3v9PvC+o2NRSil1e3TFdQ4REhLi7BAcSuvnuty5buD+\n9btbDl1Ml5VExLhKrEoplVOICOYuBq5zwhRYpZSLKl++PBEREc4OQwEBAQGEh4dn+X21JaGUumP2\nf6U6OwxFxr+Lu21J6JiEUkqpDGmSUEoplSFNEkoppTKkSUIppTLwzDPP8Oabb97RZx988EG++uqr\nLI4o++nsJqWUWwoMDGTatGm0bNnyju8xadKkLIzINblUSyIx0dkRKKXcRXJysrNDcAkulSQuXnR2\nBEopV9C/f3+OHTtGx44d8fX15f333yciIgIPDw+++uorAgICeOihhwDo0aMHpUuXpkiRIoSEhLB3\n7z9bxw0cOJDXXnsNgDVr1lC2bFk++OADSpYsib+/P19//XWm4jHGMGHCBMqXL0+pUqUYMGAAsbGx\nAFy5coV+/fpRrFgxihQpQqNGjThz5gwAX3/9NRUrVsTX15eKFSsyZ86cLPwpZY5LJYkLF5wdgVLK\nFXzzzTeUK1eOZcuWERsby8iRI1PeW7t2LWFhYaxYsQKAdu3acfjwYU6fPk3dunXp06dPhvc9efIk\ncXFxREdHM3XqVIYNG0ZMTMwt45k+fTrffPMNa9as4ciRI8TFxfHss88CMGPGDGJjY4mKiuL8+fN8\n8cUXFChQgEuXLvHcc8+xYsUKYmNjWb9+PbVr177Ln8zt0yShlHIYkax53KnrF5eJCK+//joFChQg\nX758AAwYMABvb2+8vLx47bXX2LFjB3FxceneL2/evIwdOxZPT0/atm2Lj48P+/fvv2Ucs2fP5sUX\nXyQgIABvb2/efvtt5s6di81mw8vLi3PnznHgwAFEhDp16uDj4wOAp6cnu3bt4vLly5QsWZJq1ard\n+Q/jDmmSUEo5jDFZ88hKZcqUSXlus9kYPXo0lSpVonDhwgQGBiIinD17Nt3P3nPPPXh4/PNn09vb\nm/j4+Ft+Z3R0NAEBASmvAwICuHr1KqdOnaJfv3488sgj9OzZkzJlyjB69GiSk5Px9vZm3rx5TJo0\nidKlS9OxY8dMJaSspklCKeWWJIMmSOrrs2fPZunSpaxevZqLFy8SHh6eZYf1pHbvvfem2eMqIiIC\nLy8vSpYsSZ48eRg7dix79uxh/fr1LF26lG+++QaAVq1asXLlSk6ePElQUBCDBw/O0rgyQ5OEUsot\nlSpViiNHjqS5dv0f/7i4OPLly0eRIkVISEhgzJgxGSaXu9GrVy8+/PBDwsPDiY+P59VXX6Vnz554\neHgQGhrK7t27sdls+Pj44OXlhYeHB6dPn+aHH37g0qVLeHl54ePjg6enZ5bHdiuaJJRSbmn06NH8\n5z//oWjRonzwwQfAja2L/v37U65cOfz9/alZsyZNmjS5re+4WUJJ/d6gQYPo168fzZs3p2LFinh7\ne/PJJ58A1mB4t27d8PPzo0aNGjz44IP069cPm83GBx98gL+/P8WKFWPt2rVOWbfhUrvAvviiYeJE\nZ0eilLpGd4HNOXQXWHSdhFJKZTeXShLa3aSUUtlLk4RSSqkMaZJQSimVIU0SSimlMqRJQimlVIZc\nKklcugRJSc6OQimlcg+XShK+vjoNVimlspNLJYkiRTRJKKUc69q5EdfUrFmTtWvXZqrs9Tw8PG7Y\nGsTVuNTxpUWK6LiEUsrxUm+psXv37kyXvZ33XIXDWxIi0kZEwkTkgIiMSuf9FiJyUUS22R//zuhe\nmiSUUq7EHbYscWiSEBEP4DPgEaAG0EtEqqZTdK0xpq79MSGj+xUurElCKXVr7777Lt27d09z7bnn\nnuP5558HrGNBq1evjq+vL5UqVWLy5MkZ3iswMJDVq1cDcPnyZQYMGEDRokWpWbMmW7ZsyXRMsbGx\n9O/fnxIlShAYGMibb76Z8t7hw4cJCQmhcOHClChRgl69eqW898ILL1CyZEn8/PyoVatWmuNVs4Oj\nu5saAgeNMREAIjIX6AyEXVcuU20ybUkopTKjZ8+evPHGGyQkJFCwYEFsNhvz589nyZIlAJQsWZKf\nfvqJ8uXLs27dOtq0aUPDhg1veTzo+PHjOXr0KEePHiU+Pp42bdpkOqbhw4cTFxdHeHg4Z86coXXr\n1tx7770MHDiQsWPH8sgjjxAaGkpiYiJbt24FYOXKlfz+++8cOnSIQoUKsX//fgoXLnznP5g74Ogk\n4Q9Epnp9HCtxXO9+EdkORAEvG2PSTZWaJJRyLfJ61vTJm3G3121Trlw56taty6JFi+jbty+rVq2i\nYMGCNGjQAIC2bdumlG3WrBmtW7dm3bp1t0wS8+fP54svvsDPzw8/Pz9GjBjBf/7zn1vGY7PZmDdv\nHjt37sTb25uAgABeeuklZs6cycCBA/Hy8iIiIoKoqCj8/f1Ttiz38vIiLi6OvXv30rBhQ4KCgm7r\n55AVcsLA9Z9AOWPMJRFpCywGqqRXsEgROH8+W2NTSt2F2/3jnpV69erFnDlz6Nu3L3PmzKF3794p\n7y1fvpw33niDAwcOYLPZ+PvvvwkODr7lPaOjo9Mcf5r6SNKbOXv2LElJSZQrVy7NZ6OiogCre2zs\n2LE0bNiQokWL8uKLLzJw4EAefPBBhg8fzrBhwzh27Bhdu3bl/fffTzkDOzs4OklEAeVSvS5jv5bC\nGBOf6vlyEflcRIoaY25IB7//Pp7oaBg/HkJCQggJCXFM1Eopl9e9e3dGjhxJVFQUixYtYuPGjQAk\nJibSrVs3vv32Wzp37oyHhwddunTJ1CBz6dKliYyMpFq1agBpjiS9mWLFiqW0FqpWrZryWX9/f8Dq\n/ro2LvLHH3/w8MMP06JFCypUqMDw4cMZPnw4Z8+epXv37rz33nu8/vrrGX5XaGgooaGhmYorMxyd\nJLYAlUQkADgB9AR6pS4gIiWNMafszxtiHYSUbnuhf//xzJ9vJQmllLqZYsWK0aJFCwYOHEiFChVS\numoSExNJTEykWLFieHh4sHz5clauXMl99913y3v26NGDt99+m4YNGxIfH89nn32WqVg8PDzo0aMH\nr776KjNmzODcuXN8+OGHvPLKKwAsWLCA+++/H39/fwoXLoyHhwceHh5s3boVm81G3bp1KVCgAPnz\n58fD4+bzja7/B/TNEkqmYr+rT9+CMSYZGA6sBPYAc40x+0RkiIg8bS/WTUR2i8hfwEfA4xndT8ck\nlFK3o3fv3qxatYo+ffqkXPPx8eGTTz6he/fuFC1alLlz59K5c+cM75F6rcO4ceMoV64cgYGBtGnT\nhv79+9/0+1N/9pNPPsHb25sKFSrQvHlz+vbty8CBAwHYsmULjRo1wtfXl0cffZRPPvmE8uXLExsb\ny+DBgylatCiBgYEUK1aMl19++U5/HHfEpY4v3bLFMGQI/Pmns6NRSoEeX5qT6PGlaEtCKaWym0sl\nCV1Mp5RS2culupuSkgx588LVq3CLsRulVDbQ7qacQ7ubAE9P8PGBmBhnR6KUUrmDSyUJ0HEJpZTK\nTi6VJIwxmiSUUiob5YRtOTIt5koMRYoU1oOHlMohAgIC3OLMBHeQ2S1CbpdLJYnImEiKFCmsLQml\ncojw8HBnh6AczKW6m47FHNPuJqWUykYulSQiYyN1rYRSSmUjl0oSx2KOERgIBw86OxKllModXCpJ\nRMZGUrcubNvm7EiUUip3cK0kERNJrVqwbx9cueLsaJRSyv25VJI4FnMMb2+oWBF273Z2NEop5f5c\nKklExUVhMzbq1dPtwpVSKju4VJLwy+fHqfhT1Kun4xJKKZUdXCpJlPUrmzJ4rS0JpZRyPJdKEuX8\nyhEZE0nt2rBnDyQmOjsipZRyby6VJMr6luVYzDEKFoTAQNi719kRKaWUe3O5JBEZGwmgg9dKKZUN\nXCpJlPMrl5IkdFxCKaUcz6WSRFk/q7sJtCWhlFLZwaWSxLWBa4A6dawFdVevOjkopZRyYy6VJEr7\nlObspbMkJifi4wPlyllbdCillHIMl0oSnh6elPIpRVRsFAAtW8LSpU4OSiml3JhLJQlIO3jdvz/M\nmAHGODkopZRyUy6XJFIPXjdsCB4esGGDk4NSSik35XJJIsAvgEPnDwEgAk88YbUmlFJKZT2XSxJt\nK7Vl4b6FKa/79YP58+Hvv50YlFJKuSmHJwkRaSMiYSJyQERG3aRcAxG5KiJdb3a/ZgHNiLkSw85T\nOwEoUwbq14clS7I4cKWUUo5NEiLiAXwGPALUAHqJSNUMyv0XWHGre3qIB33u68O3O79NuaZdTkop\n5RiObkk0BA4aYyKMMVeBuUDndMo9CywATmfmpn2D+zJr1yySbckAdOkCmzbB0aNZFLVSSinA8UnC\nH4hM9fq4/VoKEbkXeNQYMwmQzNy0evHqlCxYktDwUAC8veFf/4K33sqSmJVSStnlcXYAwEdA6rGK\nDBPF+PHjU543Lt6Yb3d9y0MVHgLghRegShV49VUoX94xgSqlVE4XGhpKaGholt1PjANXoolIY2C8\nMaaN/fVowBhj3klV5si1p0AxIAF42hjzw3X3MqljPRF3guqfVyfqxSi8vbwB+Pe/4fRpmDzZYVVS\nSimXIiIYYzLVS5MeR3c3bQEqiUiAiOQFegJp/vgbYyrYH4FY4xL/uj5BpKd0odI0LtOY2btmp1x7\n4QX4/nsID8/SOiilVK7l0CRhjEkGhgMrgT3AXGPMPhEZIiJPp/eR27n/mAfG8Na6t7iabG0Fe889\nMHSojk0opVRWcWh3U1a6vrvpmoe+eYi+9/VlYJ2BAJw7B5Urw44dULZsdkeplFI5S07vbnK4cS3G\nMWHdhDStiSefhPffd3JgSinlBlw+STQPaE75wuXTLK578UWYOdMaxFZKKXXnXD5JwI2tidKloWdP\n+OgjJwemlFIuzi2SRHqtiVdegS+/hIsXnRiYUkq5OLdIEgDjW4znP2v/k9KaKF8eOnSATz91blxK\nKeXK3CZJNAtoRsWiFZmx45+d/l57DT7+GI4fd2JgSinlwlx+Cmxq6yPX0/v73hx49gB5PfMCMHYs\nHDwIc+dmR5RKKZWz5PopsKk1KduEoGJBfL3965RrY8bAxo3w22/Oi0sppVyVWyUJgNdDXmfC2gkk\nJCYA1g6xH30Ew4bB1atODk4ppVyM2yWJxmUa0yygGW+t+2dvjs6dISAAJk1yYmBKKeWC3GpM4pro\nuGiCJwWz/sn1VLmnCgCbN0P37nDoEHh5OTJSpZTKOXRMIh33FrqXMQ+MYcTyEVxLLA0bWtNiFyxw\nbmxKKeVK3DJJAIxoNILI2EgWhy1Oufbyy/Dee+AijSellHI6t00SXp5eTGo/iWE/DeNU/CkA2rWD\ny5dh9WonB6eUUi7CbZMEWNt1DKoziP6L+2MzNjw8YORIqzWhlFLq1tw6SQCMDxlPQmICE9dPBKBP\nH9i1yzpvQiml1M255eym60VcjKDBlAb82PtHGvg34MMPYd06WLgwi4NUSqkc5m5nN+WKJAEwc8dM\nPtvyGRue3MCVyx5UqgTLlkGdOlkYpFJK5TA6BTaT+gT3wWZszN09lwIFYNQoGD/e2VEppVTOlmta\nEgDrItbRd1FfwoaFIckFqFQJFi+G+vWzKEillMphtCVxG5oFNKP+vfX5aONH5M8P//d/1nbiSiml\n0perWhIAh84fotHURuwbtg+/PCW47z4YN86a9aSUUu5GWxK3qVLRSnSt2pUpf04hXz6YPx+efx72\n7nV2ZEoplfPkuiQB8K8G/2Lytskk25KpVctaXPfYYxAf7+zIlFIqZ8mVSaJO6TqU8inFz4d+BmDA\nAGjaFIYMcW5cSimV02QqSYjIcyLiK5ZpIrJNRFo7OjhHGlpvKJO2/nPAxKefwtatsGSJE4NSSqkc\nJrMtiUHGmFigNVAE6Af812FRZYPHaz7OxuMbibgYAUCBAjBlinWCXUyMk4NTSqkcIrNJ4trIeDtg\npjFmT6prLsnby5u+wX2Zsm0KW6O38th3j/Hcvjq072DjlVecHZ1SSuUMmU0Sf4rISqwksUJECgG2\nzHxQRNqISJiIHBCRUem830lEdojIXyKyVURaZj78uzO0/lDeW/8eXeZ1oXm55niKJw8O+ZGffoLQ\n0OyKQimlcq5MrZMQEQ+gNnDEGHNRRIoCZYwxOzPxuQPAQ0A0sAXoaYwJS1XG2xhzyf78PmCRMaZS\nOvfKknUS19sctZlaJWuRL08+5uyaw+RtkxlV8jcGDrQSRVBQln+lUkplm+xaJ3E/sN+eIPoC/wYy\n03PfEDhojIkwxlwF5gKdUxe4liDsfICzmYwpSzT0b0i+PPkA6Fa9G4fPH6ZErW28/Ta0bg3HjmVn\nNEoplbNkNklMAi6JSC3gJeAw8E0mPucPRKZ6fdx+LQ0ReVRE9gE/ASMyGVOW8/L0YkSjEXyw4QMG\nDIAXX4SHH4bTp50VkVJKOVeeTJZLMsYYEekMfGaMmSYiT2ZVEMaYxcBiEXkAmAmk28kzPtW2rSEh\nIYSEhGRVCCmeqvsUFT6uwPHY4zz3XBlOnoR+/WD5cvDIlatKlFKuJDQ0lNAsHFTN7JjEGuBnYBDQ\nDDgN7DDG3HeLzzUGxhtj2thfjwaMMeadm3zmMNDQGHPuuusOGZNIz/M/P8+S/daCidMJp7kaW4Qg\n37p0b1qfl+5/iYJ5C2ZLHEopdbey5dAhESkF9Aa2GGPWiUg5IMQYc9MuJxHxBPZjDVyfADYDvYwx\n+1KVqWiMOWx/XheYb4ypmM69si1JXEm6wvaT2ylesDglCpZg057TdHnmL2o+8QUtgxoxoeWEbIlD\nKaXuVradTCciJYEG9pebjTGZ6qkXkTbAx1jjH9OMMf8VkSFYLYrJIvIK0B9IBBKAF4wxW9O5T7Yl\nifRMngwfTz/OyUdr8eeQPylfuLzTYlFKqczKrpZED+A9IBRrEV0z4GVjzII7/eLb5ewkYQx07gwX\ngt/g3tp7mNdtntNiUUqpzMquJLEDaHWt9SAixYFfjTG17vSLb5ezkwRYs5yC613CDKvKgl6zaBbQ\nzKnxKKXUrWTXOgmP67qXzt3GZ91GiRIwdZI3tpXvMHTZMI5eOJrynjGGNeFr2H5yuxMjVEqprJXZ\nlsR7QDAwx37pcWCnMeaGbTYcJSe0JK4ZMtSwOd8Ejvl/RO+avWkZ2JKJGyYSfjEcv/x+7HpmFx6S\n63KoUioHypaWhDHmZWAyVqIIBiZnZ4LIaT6YKFxeOZaxhcPI45GHd9e/y7AGwwh/PpxCeQuxYG+2\nDdUopZRD5bozrrPK/v3QrBksWwYNG/5z/edDP/PSype0NaGUyhEc2pIQkTgRiU3nEScisXf6pe4g\nKMg6f6Jbt7TbdjxS8RFtTSil3Ia2JO7S2LHw++/w66/g6Wld09aEUiqnyK7ZTSoD48dbezq9+eY/\n1x6p+Ag+eX34Yf8PTotLKaWygiaJu+TpCd9+C5MmwZo11jURYXDdwczaNcu5wSml1F3SJJEFSpeG\n6dOhb184az8No2u1rqw8vJK4K3HODU4ppe6CJoks0qYN9OkDnTrBhQtQtEBRmgc0T9lNVimlXJEm\niSz01lvQuDG0aAEnTkDPGj2Zu3uus8NSSqk7pkkiC3l4wMSJ8Pjj1hqKOgU7se7YOs7/fd7ZoSml\n1B3RJJHFRODVV+HJJ+GpfoVoXeERvt/7vbPDUkqpO6JJwkFGjYJChYDdPZm7Zy7JtmT2ndnH+sj1\nzg5NKaUyTRfTOdDJk1C7/t8kPHMvRq5SyqcUpxJOcWD4AUoXKu3s8JRSuYAupsvBSpWC6VMKUGjW\ndnYNiOLQiEP0qN6DmTtnOjs0pZTKFE0SDta2LfRuF8CIIX4YA4PqDOKrv77C1VpFSqncSZNENnj7\nbThzBj74AJqUbYLBsOH4hpT3P9jwATO2z3BihEoplT4dk8gmERHWluKLF8Pa5Hc4dP4QUzpN4bej\nv9FhTgeqF6/OlsFbnB2mUsrN6JiEiwgIsLYWf+wx8D/bnwX7FhBxMYJ+i/rxXbfvOHLhCJExkc4O\nUyml0tAkkY06dYIZM2D8yNIUONOUxlOa0Pu+3rSv0p6OVTqyOGyxs0NUSqk0NElks1atYNcuaJZv\nOOcPVqVzoQkAdKnahYVhC1PKGWOYt3sel5MuOytUpZTSJOEMBQrAvDfbML/DKh7tmJfFi6F1xdZs\nO7GNs5esbWQXhS2iz8I+PLH4CWzGBliJY/Svoxm6bKgzw1dK5SKaJJyoUydYvhyGDYNvvy7AwxUe\nZun+pcRcjmHE8hH81OcnouOiGfPrGGzGxojlI1h6YCmLwhbpFFqlVLbQ2U05wJEj0Lw5dBk3iwif\neZTzK8flpMtM7TSVc5fO0eSrJtxT4B4AlvdZTq0vavFz35+pWqyqkyNXSuV0dzu7SZNEDrFzJzzU\n/iLxT5ehsHch9vxrD0ULFAXg8PnDTNwwkXcefodC+QrRf1F/Hij3AE/Xe9rJUSulcjqdAusmgoNh\n0ZzCsKcHz1f5NCVBAFQsWpHP239OoXyFAGgR0IK1EWudFapSKhdxeJIQkTYiEiYiB0RkVDrv9xaR\nHfbH7yJyn6NjyqkeeABmdvuKL0Z04/xNjqBoHtCcNRFrdFxCKeVwDk0SIuIBfAY8AtQAeonI9R3p\nR4DmxphawARgiiNjyum6dYNHH4X+/cFmS79MpaKVSLIlEX4xPFtjU0rlPo5uSTQEDhpjIowxV4G5\nQOfUBYwxG40xMfaXGwF/B8eU4737rnVO9n//m/77IqJdTkqpbOHoJOEPpN5r4jg3TwJPAcsdGpEL\n8PKC776DTz+FdevSL9M8oLkmCaWUw+VxdgDXiMiDwEDggYzKjB8/PuV5SEgIISEhDo/LWfz9YepU\n6NsXtm+HIkXSvt88oDkfbPjAOcEppXKs0NBQQkNDs+x+Dp0CKyKNgfHGmDb216MBY4x557pywcD3\nQBtjzOEM7uXWU2AzMmIEnDoFc+da52dfYzM2SrxXgh1Dd+Dvm+t76JRSGcjpU2C3AJVEJEBE8gI9\ngR9SFxCRclgJol9GCSI3e/dd2LsXpk9Pe91DPGgW0IzVR1c7JzClVK7g0CRhjEkGhgMrgT3AXGPM\nPhEZIiLXVoKNBYoCn4vIXyKy2ZExuZr8+WHOHBgzBl55BS5d+ue9IfWGMGbVGE7Fn0q5ZjM2fjzw\nIxEXI5wQrVLK3eiKaxdx6hS88AJs3AhffQXXhmPGrh7L75G/80u/X0i2JTPoh0FsO7GNc5fOUcqn\nFJ2DOtM3uC9BxYKcGr9Syjl0W45cZvlyeOIJq3Xx0EOQbEum7ay2BN0TxJ4zeyicvzCzus4ir2de\nNkVtYsHeBczeNZvyhcszoeUEHq7wsLOroJTKRpokcqG1a60T7pYsgSZN4Oyls9SfXJ9OQZ348JEP\n8fTwTFM+yZbEnF1zeH3N6xx49gAeoruxKJVbaJLIpVasgH794OefoW5dKxHk8ch4RrMxhjpf1uHd\nVu/SumLrbIxUKeVMOX12k3KQRx6BL7+E9u2t2U83SxBg/YfyTP1nmLR1UjZFqJRyB5okXFiXLvDe\ne9C6NRzOxOThPsF9WBO+huOxxx0fnFLKLWiScHF9+8Jrr8HDD0N09M3L+uT1oVfNXkzdNjV7glNK\nuTwdk3ATb7wBq1ZZjzw36XnadWoXbWa1Ify5cLw8vbIvQKWUU+iYhALg1VchXz5Itb1Vuu4reR+B\nhQNZsn9JtsSllHJtmiTchKcnfPstfP21NfPpZkY1HcXra17HZjI4sEIppew0SbiREiWsRNG/P0ye\nDImJ6ZfrUKUDPnl9mLt7bpZ+/69HfmXgkoFZek+llHNpknAzISGweDF8/z1UrgzffHNjGRHhzZZv\nMi50HFeTr2bZd3+y6RNWHl6ZZfdTSjmfJgk3dP/9VpfTvHnWzKfZs28s0zKwJQF+AXy9/WsAEhIT\n2Hlq5x1/Z1RsFOuOrSPuShxnL5294/sopXIWnd3k5nbutKbHLlliJY/UNh3fxGPfPUaTsk1YcXgF\ngvBNl2/oFNTptr/nzbVvcizmGHvP7uWNkDd4MPDBLKqBUupu6OwmdVPBwdZZFI89BuHhad9rVKYR\nwxsOp1WFVhwZcYRlvZcxZNmQNFuPZ4bN2Jj21zQG1xtMcIngu2qRKKVyFk0SuUD79tYU2ebNYcOG\ntO+NfmA0g+sN5h7ve3ig3AMMqj2IJ394kttpta0+uhrffL7UK12P4JKaJJRyJ5okcolhw+B//4PO\nneGTTyCjHDAuZBwn4k/wxdYvMn3vKdumMLjuYETEShKnNUko5S50TCKXOXIEune3nj/1FPTqBYUL\npy0TdjaMZtObsbzPcurfW/+m9zuTcIbKn1bm6HNHKVKgCHFX4ig1sRSxo2Nv2LJcKZX9dExC3ZYK\nFWDzZnilkvscAAAZ5klEQVTrLfjtNwgMhKVL05apWqwqX3b4kq7zuqaMT1z4+wJDlw1lTfiaNGW/\n3v41Xap1oUiBIgAUyleIUj6lOHT+ULbURynlWJokciFPT2ur8e++g5Ur4cknYd26tGW6VuvKgNoD\n6D6/O8sPLqfWF7U4dP4Qo1eNThmvsBkbX/75JUPrDU3zWR2XUMp9aJLI5Ro0gFmzrNlPO3akfW98\nyHgK5y/M4KWDmdZpGiv6ruDC3xf4Lfw3AFYdWUWhfIVo6N8wzed0hpNS7kOThKJVK/jsM2jXDiIi\n/rnuIR4s6LGAA88eoFXFVnh6eDL6gdFMWDsBgC/+/IKh9YYikra7UwevlXIfmiQUAD16wEsvQceO\nEBv7z/W8nnnx9vJOed3nvj4cuXCEhfsW8tvR3+h9X+8b7qXdTUq5D00SKsULL0DTptCzJyQlpV/G\ny9OLV5q+Qp+FfXi8xuMUylfohjIVi1bkTMIZYi7HODhipZSjaZJQKUSsNRTJyfD003D5cvrlBtUZ\nRMUiFXmmwTPpvu8hHtQsUZPdp3c7MFqlVHbQJKHS8PKyZj3FxED9+vDXXzeWyZ8nP7ue2UVwyeAM\n7xNcMpjtJ7c7MFKlVHbQJKFu4OcHCxbAmDHWVNl//xvi49OWuX6w+npdqnZh4oaJ2uWklIvTJKHS\nJQJ9+lgtifBwqFrVOvXOlsnD7NpWbssjFR/hmR+fua19oJRSOYtuy6EyZdMmGD4c7rsPpk2zksit\nXLp6iQZTGvBKk1d4ovYTjg9SKXWDHL8th4i0EZEwETkgIqPSeT9IRNaLyGURedHR8ag706iRtY3H\n7t1W91NmeHt5M/exuYz8ZSQ/H/pZWxRKuSCHtiRExAM4ADwERANbgJ7GmLBUZYoBAcCjwAVjzAcZ\n3EtbEjnAmTPwwANWq+LZZzP3mZ8O/sTIlSPxyevD2OZj6RjU0bFBKqVS5PSWREPgoDEmwhhzFZgL\ndE5dwBhz1hjzJ5DBzHyVkxQvbh2N+v770LIlTJkC58/f/DPtKrdj9792M/qB0Ty/4nmmbpuaPcEq\npe6ao5OEPxCZ6vVx+zXlwsqXh/37rZbEypVQsSIMGQIHDmT8GQ/xoGu1rvzU+yf+b9X/sSVqS7bF\nq5S6czq7Sd2R/PmhSxeYP99KGKVKWd1QHTrAzJnWOov0BBUL4ssOX9JtfjfOJJzJ8P7bTmxj2I/D\nOHz+sINqoJTKjDwOvn8UUC7V6zL2a3dk/PjxKc9DQkIICQm501upLFSiBLz+OrzyCixcaCWOYcOg\nb1947z0oWDBt+S7VurA5ajPNpjfDN58vx2OPk2RLokaJGtQoXoMdp3YQcTGCKvdU4f317zOpwyTn\nVEwpFxQaGkpoaGiW3c/RA9eewH6sgesTwGaglzFmXzplxwHxxpiJGdxLB65dyIULMGIEbN0Ks2dD\nnTpp30+2JfPLkV8onL8wZXzL4Cme7D69m12ndxHgF0Dnqp05k3CG6p9X58iIIymHGimlbs/dDlw7\nfJ2EiLQBPsbq2ppmjPmviAwBjDFmsoiUBLYChQAbEA9UN8bEX3cfTRIuaPZseP55mDTJOrPidvVd\n2Jc6perwUpOXsj44pXKBHJ8ksoomCde1fTu0bm11RT3wwO19dnPUZh5f8DiHnj2kZ2YrdQdy+hRY\npahdG779Frp1g7CwW5dPraF/Q0oWLMmyA8scE5xS6qY0Sahs0bo1/Pe/0LYtLFuW+T2gAEY0GsFH\nmz4i2ZbsuACVUunS7iaVrRYuhDfesA41GjbMWmNRogRUqgQ+Pul/JjE5kZYzWrL/3H46VOlAl6pd\naFe5HXk80p+cl2xL1q4ppex0TEK5HGOsRXizZkF0NJw6Za2r+OknqFkz48+FXwxn6f6lzNk9h8jY\nSIbUG0L36t0JLBJIXs+8HDx3kE82fcLXO75mxqMz6Fqta/ZVSqkcSpOEcguzZlnHp86fDy1a3Lr8\n9pPb+XzL5/x65Fei4qIo5l2Mq8lXGVx3MHVL12XEzyPY/cxunTqrcj1NEspt/Por9OoFTz0FDz4I\n998PhW48QvsGSbYkomKjKFGwBAW8CgAw7MdhJCYnMqXTFADWRaxj1dFVvNbiNTxEh+JU7qFJQrmV\nffusVsXatbBtG7RvD6NH37gY71Zir8RS4/MaTOs0jdDwUKZvn45/IX8al2nMp20/veXJekpdb9uJ\nbUz5c4rL7QCgU2CVW6lWDSZMsJLEiRPQsCF07GjNjpo2DU6ezNx9fPP58r92/6PNt23YeWon24ds\nZ1X/VayPXM+40HFcTrrMpC2TqPxpZdrOasvOUzsdWzHl0vae2Uu7We2YvXs2+8/ud3Y42UpbEirH\nu3LFmhW1eLE14N2gAXz3HRQufOvP/nXiL2qXqp3ScjidcJpm05tx7tI5GpVpxKimo9hxcgcT1k2g\nXeV2PBz4MIFFAgksHEgpn1La4lAcuXCE5tOb8/ZDb7M1eivFvIsxtsVYZ4eVadrdpHKVxEQYORI2\nb7bOtfDzu/17nL10ltMJp6levHrKtYuXL/L5ls/ZeWon4RfDOXLhCAlXE6hQpAJlfcvik9cHn7w+\ntKvcjm7Vu2UcX3IieT3z3knV1G2KjImk/+L+zOs2jxIFSzjkO/449ge9F/ZmdNPRPNPgGf449gdD\nlg1h9792O+T7HEGThMp1jLHOsvjzTytR+Po65ntir8Ry5MIRjsceJz4xntgrsbzzxzs8GvQo77R6\nJ806jbgrcby08iW++usr2lRqw1N1n6J95fZ4eXo5JrgcZNmBZdQoXoPAIoEO+w5jTJpWnTGGTnM7\ncfDcQYKKBbH48cVZ2uq7knSFcaHjmLFjBpPaT+LRqo8CYDM2Aj4K4Oc+P1OjRI0s+75rLl29xMbj\nG2kZ2DLL7qljEirXEYFPP4X69a2T8kqXhurVrdXcI0fC9Olw7tzdf49vPl9ql6pNhyod6FmzJ0/X\ne5otg7ew58weWs9szfw981l5eCUL9y2k1he1sBkbkS9E0r16dyZumEiFTyrw8caPuXT1EgAX/r7A\nz4d+JvxieLrfdznpMuNDx/PsT89yOenyDe9fSbrCnF1z6LuwL+/8/g6bjm/iavLVG8qdu3SO+MT4\nG64DRMdF0/v73ni/6U21/1Wj/ez2fLn1S2zmNpbAp/Jn9J/0X9SfxtMas+LQikx/zhjD9L+mE3Ex\n4pZlz146S6OpjXhi8RMp9f1uz3ccvXCUrU9v5VjMMab9Ne2m90hMTmTVkVUZnmFiMzZ+OfwLo34Z\nRbtZ7Qj4KICws2HsGLojJUGAdXhW9+rd+W7PdzfcI9mWzNqItcRdibtlna636sgqen/fm3sn3ss7\nf7yT7u/VWbQloVza339b25KfPw9Hj8LevdasqF9+gd69rbUXFSum/YzNZiWaO/2HZ7ItmYkbJrI5\najMxV2K4nHSZ0U1H075K+zTltkZv5c11b7IhcgOlfEpx+MJh6pauy94ze6lctDI9avSgRvEalC9c\nnuOxx3nmx2eoVrwaeTzyEB0XzeLHF1O8YHEOnT/EpC2T+GbnN9QqWYvHqj1G2NkwQiNCCb8YTrNy\nzWgZ2BIP8WBR2CK2n9xOsi2ZoGJBNCnThDK+ZShSoAhnEs7w4cYPebre07zQ+AVOJZzi4LmDvLv+\nXfLnyc/UjlMp41uG9ZHr2Ry1me41ulOhSIUMfw6Xky5Tb3I9Xm32KmV8y9BzQU+GNRjGqAdGZbga\nHqx/LQ9cMpCdp3Zy4e8LzOwyk1YVW2GM4Zcjv7A5ajP9a/WnnF85TsSdoNXMVrSr3I6ws2HWtOaO\nU2g0tRHf9/ie+8vez57Te2jxdQtCB4QSHRfNikMruHj5ItWKV6PKPVX449gffL3ja8r4luHw+cM0\nLtOYTkGdKO5dHG8vbw6dP8TnWz8nf578dKvWjeCSwQSXDKZ84fLptk42Hd/EgCUD2PuvvYgI5/8+\nz9RtU5m0dRKe4klez7wsenwRQcWCbvhszOUYriRfSekeS7Yl8+/V/2bO7jmMajqKx6o/luVdZ9rd\npFQ6TpyAzz6DL7+E4GB44gnrf+fNszYbLFLEGgi/PoE4wr4z+4i5EkPd0nXJ65mXq8lX+fXIrywO\nW8zB8weJiInAGMP7rd+na7Wu2IyNsavHMnfPXILuCWJr9FYG1h7I0/WepmLRtAGfvXSW347+xqqj\nq0iyJfFo1Ud5uMLDCMKfJ/5k4/GNnIw/yYW/LwDwctOXqXJPlTT3SLYl89HGj3jr97e4mnyVasWr\nEVwimIVhC3mqzlO80vQVjl48yobIDVxJvkK/4H6U9CnJyJUjiYiJ4Ltu3yEiRMVG0W9RP85cOsPE\n1hNpXbF1yncYY4i5EsPRC0d5etnTBN0TxNROU9l0fBO9vu9Fl6pdWHdsHQZD07JNmb93Pi0DW7L9\n5HYG1BrAq81fJcmWxNBlQ5mzew6Dag/i03afptz/s82f8cKKF2jo35A2FdtQomAJ9p3dx/5z+6lZ\nvCaD6w2myj1VSEhMYNmBZaw8vJLYxFgSEhMo5l2Mp+s9TdOyTTPVZWWMIfDjQBY9voiNxzcyLnQc\nbSu3ZXiD4TTwb8DUbVP5v1X/x0dtPqJikYrEXInh8PnDLN6/mA2RG/AQD1oGtmRQnUFM2jqJS1cv\n8V237yhesPjd/GeWIU0SSt3ElSuwdCnMmAF79kD37tC/P6xZY52m9+230KqVs6NM38J9C4m7Esfj\nNR8nf578Dv++k/EnyeeZL2WVenRcNP9e/W++2fENVYtV5f4y92MzNhaGLaRZuWZsjd7KjqE70vxx\nM8bww/4fGPnLSArltVZCnvv7HGcSzuDl6UVpn9IMqjOIl5u8nPIH+Xjscd774z3aVW5H64qtERHi\nrsQxY8cMfPP50r9W/zT3n7N7Dh2rdKRQvrQrLa8mX822MaBXfnmFSVsnUbd0XT5t+ynBJYPTvL85\najPPLn8Wm7Hhl88Pf19/OlTuQJtKbTAYZu+azfTt07m/zP281+o9h8atSUKpO7RmDfTsaT3GjLE2\nGrTZrF1qV66EsWOhZElnR+l8SbakNN1HFy9f5Nud3xJcMpjmAc3T/UxiciLrI9dTKG8hihYoSvGC\nxfHJm8EOji4oOi6ardFb6VilY46fJq1JQqm7cPIkvPWWtcq7Z08rceTNC40awY8/Wusz6td3dpRK\n3TlNEkplgYgImDzZ2lywVStrUHvhQhgyBJ57DsqWhYIFrS3Ng4PBQ+cFKhehSUIpB9q1yxr8jouD\n+HjrdWystU1I2bLg6Qn581uv69W78xlTSjmKJgmlstnRo9YU2zNnIDnZOgtjyRLIk8eadtu5s9Xa\n0IShcgJNEkrlAMbApk0wZ441myo5GTp1gh49oGlTa0D8t9+sabfVq0OfPv/sPZWcbLVUMrMXlVK3\nS5OEUjmMMdaW5wsXWhsRnj8PV69CQAB07Qrbt8PPP8NDD1mn8u3YYSWRqlWhXTsICbFO6CtRQlsj\n6u5pklAqh9u/3+qKSr1w78wZ67jWcuWsszIKFoSNG60ZVevXW2s6RKytRp57zpphdeoUfPihte1I\ncjLky2eNi4waBY8+qglFpU+ThFJuyBhr1fjs2dY+VUWLQni41U313HPWivHLl61NDsePt2Zbdehg\nTemNjrZmYfXoYU3l1ZlYuZsmCaXcXFISrF4NtWqlv7jPZrMGzrdtg3vvtTY83L7dOi88NtYaTH/i\nCWssJDbWarEcPw5BQVCjxo1jIcZYCejiRWtfrMRE6zCofPmyp74qa2mSUEplaPdumDnT2n4kTx44\ne9aaqhsQAAcOWBsiJiVZrQ0PD2vs5MoVKyEULmy1WDw8rFbMffdZScXHx+oeMwYSEqxNFh97DNq0\n+ed7z52zZoAVLGjdp3x5q2tMZT9NEkqpW0pOhrAwqxsqdYvAGOuPvDFWi8TLy1pxfn0XVUKC1bUV\nFmY9T0iwxkAKFrT+93//g8qV4dVXrVbN1KnwwAPW9164YCWkokXh4YehcWOrVVOlCkRFwV9/Wff1\n87NaSpUrW6cP6hhL1tAkoZRyusREa9fdjz+2pv6+/LI1KH+NzQY7d8Kvv1rdYnv3WgP6pUtD3bpW\nd1Z8vDU4/+efVqtn6FBrTOXYMaslExZmDegfOACFClktk/LlrZZRgwZWAoyJsWaTxcZa90tIgMBA\na+A/v32PRGOs97y9rcWQ7i7HJwkRaQN8hHXA0TRjzDvplPkEaAskAAOMMdvTKaNJQqlcwBhYuxYm\nTbISQvnyVvfYtTGUoCDrj/yxY3DkCGzdClu2WInkWheZn5/VLebtbd1j3z7rs3Fx1hYsxljdagUK\nWGULFbJOOPT0tLrcrl61viM21vqMh4fVysqT55+zSHx9rdZQUJA1FuTjY93n2sPHxyp/7Z7Hj1sP\nHx9rkkFx++a5CQlW/GfPWs8TE2Hw4Kz7eeboJCEiHsAB4CEgGtgC9DTGhKUq0xYYboxpLyKNgI+N\nMY3TuZdbJ4nQ0FBCQkKcHYbDaP1clzvULT7eGsz387MSjq+v1bqJj4cVK0KpWTOE2Fire8zLy3r4\n+PyTbGw26w99UpKVYIz5pxtt/36rBRQfbyWUa4/4eKt8crKVLPz9rdbP6dPWeE1wsPX+zp3W89Kl\nre47X1+rVZZV7jZJZHx8VNZoCBw0xkQAiMhcoDMQlqpMZ+AbAGPMJhHxE5GSxphTDo4tR3GH/yPe\njNbPdblD3Xx8rDGS1Dw8rD/Ie/aE0r17yG3fs0QJqxXRsePtx3P5srUCP18+a4zG2/v275FdHJ0k\n/IHIVK+PYyWOm5WJsl/LVUlCKZV75M9vLZR0BbrMRimlVIYcPSbRGBhvjGljfz0aMKkHr0XkC+A3\nY8w8++swoMX13U0i4r4DEkop5UA5eUxiC1BJRAKAE0BPoNd1ZX4AhgHz7EnlYnrjEXdTSaWUUnfG\noUnCGJMsIsOBlfwzBXafiAyx3jaTjTE/iUg7ETmENQV2oCNjUkoplXkus5hOKaVU9nOJgWsRaSMi\nYSJyQERGOTueuyEiZURktYjsEZFdIjLCfr2IiKwUkf0iskJE/Jwd690QEQ8R2SYiP9hfu0397NO0\n54vIPvvvsZGb1W+MvV47RWSWiOR15fqJyDQROSUiO1Ndy7A+9voftP9+Wzsn6szLoH7v2uPfLiLf\ni4hvqvduq345PknYF+R9BjwC1AB6iUhV50Z1V5KAF40xNYD7gWH2+owGfjXGBAGrgTFOjDErPAfs\nTfXaner3MfCTMaYaUAtr3Y9b1M8+fjgYqGOMCcbqku6Fa9dvOtbfj9TSrY+IVAd6ANWwdoH4XCTH\n7yKVXv1WAjWMMbWBg9xF/XJ8kiDVgjxjzFXg2oI8l2SMOXlt2xFjTDywDyiDVacZ9mIzgEedE+Hd\nE5EyQDtgaqrLblE/+7/ImhljpgMYY5KMMTG4Sf2AWCARKCgieYACWGuXXLZ+xpjfgQvXXc6oPp2A\nufbfazjWH9jr13blKOnVzxjzqzHGZn+5EetvDNxB/VwhSaS3IM/fSbFkKREpD9TG+iWmrDI3xpwE\nSjgvsrv2IfAykHrAy13qFwicFZHp9u60ySLijZvUzxhzAZgIHMNKDjHGmF9xk/qlUiKD+mS0uNeV\nDQJ+sj+/7fq5QpJwSyLiAywAnrO3KK6fQeCSMwpEpD1wyt5aulkz1iXrh9X9Uhf4nzGmLtaMvNG4\nz++vAvACEADci9Wi6IOb1O8m3K0+AIjIq8BVY8ycO72HKySJKCDVpsOUsV9zWfZm/AJgpjFmif3y\nKREpaX+/FHDaWfHdpaZAJxE5AswBWorITOCkm9TvOBBpjNlqf/09VtJwl99ffeAPY8x5Y0wysAho\ngvvU75qM6hMFpD4eyWX/3ojIAKxu396pLt92/VwhSaQsyBORvFgL8n5wckx36ytgrzHm41TXfgAG\n2J8/ASy5/kOuwBjzf8aYcsaYCli/q9XGmH7AUtyjfqeASBGpYr/0ELAHN/n9AfuBxiKS3z6g+RDW\nBARXr5+QtmWbUX1+AHraZ3QFApWAzdkV5F1IUz+xjmh4GehkjLmSqtzt188Yk+MfQBus/3gPAqOd\nHc9d1qUpkAxsB/4CttnrVxT41V7PlUBhZ8eaBXVtAfxgf+429cOa0bTF/jtcCPi5Wf1exkp8O7EG\ndb1cuX7AbKyjCq5gjbUMBIpkVB+smUCHsCaVtHZ2/HdYv4NAhP3vyzbg8zutny6mU0oplSFX6G5S\nSinlJJoklFJKZUiThFJKqQxpklBKKZUhTRJKKaUypElCKaVUhjRJKJUNRKSFiCx1dhxK3S5NEkpl\nH12UpFyOJgmlUhGRPiKyyb7D6yT74UlxIvKBiOwWkV9E5B572doisiHVwS5+9usV7eW2i8hW+/YH\nAIVSHVY002mVVOo2aJJQys5++NPjQBNj7fBqA/oA3sBmY0xNYC0wzv6RGcDLxjrYZXeq67OAT+3X\nmwAn7NdrAyOA6kBFEWni+FopdXfyODsApXKQh7B2dN1i39wuP3AKK1l8Zy/zLXDtOEg/Yx34AlbC\n+M6+Bby/MeYHAGNMIoD98K/NxpgT9tfbgfLA+myol1J3TJOEUv8QYIYx5tU0F0XGXlfOpCp/O1Lv\nxpmM/v9PuQDtblLqH6uAbiJSHEBEiohIOcAT6GYv0wf43RgTC5wXkab26/2ANcY6QCpSRDrb75FX\nRApkay2UykL6Lxml7Iwx+0Tk38BKEfHAOut5ONbpcw3tLYpTWOMWYJ1D8KU9CRzB2qIZrIQxWUTe\nsN+je3pf57iaKJV1dKtwpW5BROKMMYWcHYdSzqDdTUrdmv5LSuVa2pJQSimVIW1JKKWUypAmCaWU\nUhnSJKGUUipDmiSUUkplSJOEUkqpDGmSUEoplaH/B+M99AM7jl7IAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb61c5e4850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this is the plot on cylinder_small1 with regularization\n",
    "plot_loss(cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8017638036809815"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# this is the score on cylinder_small1 without regularization\n",
    "cnn.score({'image_input': X_test['image'],\n",
    "           'prob_input': X_test['prob'],\n",
    "           'binary_input': X_test['merged_array'],\n",
    "           'border_input': X_test['border_overlap']},\n",
    "           y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAEPCAYAAAC3NDh4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npZRSufp/5VSZxzSRg+oAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f42bc456b10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this is the plot on cylinder_small1 without regularization\n",
    "plot_loss(cnn)"
   ]
  },
  {
   "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
}