keras/convmnist/keras_cnn_mnist_v1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN on mnist with convolution visualization\n",
    "\n",
    "This is heavily based on keras' example : https://github.com/fchollet/keras/blob/master/examples/mnist_cnn.py\n",
    "\n",
    "This is for the **keras v1** api"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "os.environ['THEANO_FLAGS']='mode=FAST_RUN,device=gpu,floatX=float32'\n",
    "# This gives a 'perform' error in compile\n",
    "#os.environ['THEANO_FLAGS']='mode=FAST_COMPILE,device=gpu1,floatX=float32'\n",
    "#os.environ['THEANO_FLAGS']='device=gpu0'\n",
    "\n",
    "#os.environ['KERAS_BACKEND'] = 'tensorflow'\n",
    "#os.environ['THEANO_FLAGS']='mode=FAST_COMPILE,device=cpu,floatX=float32'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/julien/dev/venv/local/lib/python2.7/site-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.\n",
      "  warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')\n",
      "Using Theano backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Couldn't import dot_parser, loading of dot files will not be possible.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 0: GeForce GTX 960 (CNMeM is disabled, cuDNN 5004)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\n    Train a simple convnet on the MNIST dataset.\\n\\n    Run on GPU: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python mnist_cnn.py\\n\\n    Get to 99.25% test accuracy after 12 epochs (there is still a lot of margin for parameter tuning).\\n    16 seconds per epoch on a GRID K520 GPU.\\n'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning) \n",
    "\n",
    "from __future__ import absolute_import\n",
    "from __future__ import print_function\n",
    "import pylab as pl\n",
    "import matplotlib.cm as cm\n",
    "import numpy as np\n",
    "np.random.seed(1337) # for reproducibility\n",
    "\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers.convolutional import Convolution2D, MaxPooling2D\n",
    "from keras.utils import np_utils\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "'''\n",
    "    Train a simple convnet on the MNIST dataset.\n",
    "\n",
    "    Run on GPU: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python mnist_cnn.py\n",
    "\n",
    "    Get to 99.25% test accuracy after 12 epochs (there is still a lot of margin for parameter tuning).\n",
    "    16 seconds per epoch on a GRID K520 GPU.\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gpu\n"
     ]
    }
   ],
   "source": [
    "import theano\n",
    "print(theano.config.device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.set_printoptions(precision=5, suppress=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train shape: (60000, 1, 28, 28)\n",
      "60000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "nb_classes = 10\n",
    "\n",
    "# the data, shuffled and split between tran and test sets\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "X_train = X_train.reshape(X_train.shape[0], 1, 28, 28)\n",
    "X_test = X_test.reshape(X_test.shape[0], 1, 28, 28)\n",
    "X_train = X_train.astype(\"float32\")\n",
    "X_test = X_test.astype(\"float32\")\n",
    "X_train /= 255\n",
    "X_test /= 255\n",
    "print('X_train shape:', X_train.shape)\n",
    "print(X_train.shape[0], 'train samples')\n",
    "print(X_test.shape[0], 'test samples')\n",
    "\n",
    "# convert class vectors to binary class matrices\n",
    "Y_train = np_utils.to_categorical(y_train, nb_classes)\n",
    "Y_test = np_utils.to_categorical(y_test, nb_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label :  [ 0.  0.  0.  0.  0.  1.  0.  0.  0.  0.]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f615ae94710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = 4600\n",
    "pl.imshow(X_train[i, 0], interpolation='nearest', cmap=cm.binary)\n",
    "print(\"label : \", Y_train[i,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Convolution2D(32, 3, 3, border_mode='valid', input_shape=X_train.shape[1:]))\n",
    "# The Dropout is not in the original keras example, it's just here to demonstrate how to\n",
    "# correctly handle train/predict phase difference when visualizing convolutions below\n",
    "model.add(Dropout(0.1))\n",
    "convout1 = Activation('relu')\n",
    "model.add(convout1)\n",
    "model.add(Convolution2D(32, 3, 3))\n",
    "\n",
    "convout2 = Activation('relu')\n",
    "model.add(convout2)\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "model.add(Dense(nb_classes))\n",
    "model.add(Activation('softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adadelta', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Uncomment to visualize\n",
    "#from IPython.display import SVG\n",
    "#from keras.utils.visualize_util import model_to_dot\n",
    "#SVG(model_to_dot(model, show_shapes=True).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.2666 - acc: 0.9190 - val_loss: 0.0604 - val_acc: 0.9814\n",
      "Epoch 2/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0976 - acc: 0.9708 - val_loss: 0.0402 - val_acc: 0.9874\n",
      "Epoch 3/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0719 - acc: 0.9784 - val_loss: 0.0353 - val_acc: 0.9884\n",
      "Epoch 4/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0589 - acc: 0.9820 - val_loss: 0.0328 - val_acc: 0.9895\n",
      "Epoch 5/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0507 - acc: 0.9847 - val_loss: 0.0268 - val_acc: 0.9913\n",
      "Epoch 6/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0474 - acc: 0.9854 - val_loss: 0.0291 - val_acc: 0.9909\n",
      "Epoch 7/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0412 - acc: 0.9871 - val_loss: 0.0295 - val_acc: 0.9906\n",
      "Epoch 8/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0373 - acc: 0.9882 - val_loss: 0.0313 - val_acc: 0.9904\n",
      "Epoch 9/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0339 - acc: 0.9889 - val_loss: 0.0276 - val_acc: 0.9915\n",
      "Epoch 10/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0333 - acc: 0.9896 - val_loss: 0.0280 - val_acc: 0.9908\n",
      "Epoch 11/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0310 - acc: 0.9905 - val_loss: 0.0309 - val_acc: 0.9906\n",
      "Epoch 12/12\n",
      "60000/60000 [==============================] - 8s - loss: 0.0289 - acc: 0.9910 - val_loss: 0.0265 - val_acc: 0.9923\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/julien/dev/venv/local/lib/python2.7/site-packages/keras/models.py:426: UserWarning: The \"show_accuracy\" argument is deprecated, instead you should pass the \"accuracy\" metric to the model at compile time:\n",
      "`model.compile(optimizer, loss, metrics=[\"accuracy\"])`\n",
      "  warnings.warn('The \"show_accuracy\" argument is deprecated, '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test score: 0.0265096118289\n",
      "Test accuracy: 0.9923\n"
     ]
    }
   ],
   "source": [
    "WEIGHTS_FNAME = 'mnist_cnn_weights_v1.hdf'\n",
    "if True and os.path.exists(WEIGHTS_FNAME):\n",
    "    # Just change the True to false to force re-training\n",
    "    print('Loading existing weights')\n",
    "    model.load_weights(WEIGHTS_FNAME)\n",
    "else:\n",
    "    batch_size = 128\n",
    "    nb_epoch = 12\n",
    "    model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,\n",
    "              verbose=1, validation_data=(X_test, Y_test))\n",
    "    model.save_weights(WEIGHTS_FNAME, overwrite=True)\n",
    "score = model.evaluate(X_test, Y_test, show_accuracy=True, verbose=0)\n",
    "print('Test score:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  1.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]]\n",
      "[[ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  1.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]]\n"
     ]
    }
   ],
   "source": [
    "print(model.predict(X_test[1:5]))\n",
    "print(Y_test[1:5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "Y_pred = model.predict(X_test)\n",
    "# Convert one-hot to index\n",
    "y_pred = np.argmax(Y_pred, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.99      1.00      0.99       980\n",
      "          1       0.99      1.00      1.00      1135\n",
      "          2       0.99      0.99      0.99      1032\n",
      "          3       0.99      0.99      0.99      1010\n",
      "          4       0.99      0.99      0.99       982\n",
      "          5       0.99      0.99      0.99       892\n",
      "          6       0.99      0.99      0.99       958\n",
      "          7       0.99      0.99      0.99      1028\n",
      "          8       0.99      0.99      0.99       974\n",
      "          9       0.99      0.99      0.99      1009\n",
      "\n",
      "avg / total       0.99      0.99      0.99     10000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convolution visualizations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from keras import backend as K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# K.learning_phase() is a flag that indicates if the network is in training or\n",
    "# predict phase. It allow layer (e.g. Dropout) to only be applied during training\n",
    "inputs = [K.learning_phase()] + model.inputs\n",
    "\n",
    "_convout1_f = K.function(inputs, [convout1.output])\n",
    "def convout1_f(X):\n",
    "    # The [0] is to disable the training phase flag\n",
    "    return _convout1_f([0] + [X])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# utility functions\n",
    "from mpl_toolkits.axes_grid1 import make_axes_locatable\n",
    "\n",
    "def nice_imshow(ax, data, vmin=None, vmax=None, cmap=None):\n",
    "    \"\"\"Wrapper around pl.imshow\"\"\"\n",
    "    if cmap is None:\n",
    "        cmap = cm.jet\n",
    "    if vmin is None:\n",
    "        vmin = data.min()\n",
    "    if vmax is None:\n",
    "        vmax = data.max()\n",
    "    divider = make_axes_locatable(ax)\n",
    "    cax = divider.append_axes(\"right\", size=\"5%\", pad=0.05)\n",
    "    im = ax.imshow(data, vmin=vmin, vmax=vmax, interpolation='nearest', cmap=cmap)\n",
    "    pl.colorbar(im, cax=cax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEKCAYAAABdduPuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6105a39090>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = 4600\n",
    "\n",
    "# Visualize the first layer of convolutions on an input image\n",
    "X = X_test[i:i+1]\n",
    "\n",
    "pl.figure()\n",
    "pl.title('input')\n",
    "nice_imshow(pl.gca(), np.squeeze(X), vmin=0, vmax=1, cmap=cm.binary)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy.ma as ma\n",
    "def make_mosaic(imgs, nrows, ncols, border=1):\n",
    "    \"\"\"\n",
    "    Given a set of images with all the same shape, makes a\n",
    "    mosaic with nrows and ncols\n",
    "    \"\"\"\n",
    "    nimgs = imgs.shape[0]\n",
    "    imshape = imgs.shape[1:]\n",
    "    \n",
    "    mosaic = ma.masked_all((nrows * imshape[0] + (nrows - 1) * border,\n",
    "                            ncols * imshape[1] + (ncols - 1) * border),\n",
    "                            dtype=np.float32)\n",
    "    \n",
    "    paddedh = imshape[0] + border\n",
    "    paddedw = imshape[1] + border\n",
    "    for i in xrange(nimgs):\n",
    "        row = int(np.floor(i / ncols))\n",
    "        col = i % ncols\n",
    "        \n",
    "        mosaic[row * paddedh:row * paddedh + imshape[0],\n",
    "               col * paddedw:col * paddedw + imshape[1]] = imgs[i]\n",
    "    return mosaic\n",
    "\n",
    "#pl.imshow(make_mosaic(np.random.random((9, 10, 10)), 3, 3, border=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "W shape :  (32, 3, 3)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA4YAAANFCAYAAAAnD+IXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3XvUtXVZL/rvxXIheCJAQBJBEcQyNiQHKTMI87DMxIGm\nYIamORzbYK1trjLMFLOlYupeo8xMZLsiDVIDRZcBuhWSwuKQBwQE5CDIC6JiHtimwm//8Uxnj6/v\n+X7e5543v89njGcwD/ec9/cBBnBxfe85q7UWAAAA+rXN2AEAAAAYl8EQAACgcwZDAACAzhkMAQAA\nOmcwBAAA6JzBEAAAoHMGQwAmr6r+oqr+YBOPfVdV/dHWzgQAU2IwBGBFVdWDquqDVfXlqrq7qvbc\n2udsrf2frbX/sRLvNcu890q8FwBMhcEQgJV2d5K/T3J0kjZyli0xxcwAMIjBEOAerqr2qKq/q6qv\nVNXtVfWns8erql5ZVTdU1a1V9b+q6gGz5/aabc6Oq6obZ699xey53avqzqr6iWXn+NnZe/+n1tpX\nWmtvT3JJktpItudX1dnL7l9TVX+77P6Xqur/mN1+ZFWdV1Vfq6orq+rXlh33I/XQqvq9qrqlqm6u\nqheuYwu4U1V9uKq+WVUXVdXDZq+7YJb5s7Pnfq2qdq6qD1XVHbNzX7AFfxkAYKEZDAHuwapqmyQf\nTnJ9kj2TPDjJGbOnfzPJcUkOT7J3kvsneetab/HYJPsm+eUkr6qq/Vpra5L8U5JnLDvu2CTva63d\ntZkRL0jyC7Osuyf5z0l+bnZ/7yT3ba19tqruk+S8JO9O8sAkxyR5W1U9ch2/85OT/F9JjkyyT5Ij\n8uNbwGcneXWSn0jyxST/I0laa4fPnt+/tfaA1tr7krwsyU1Jdk6ya5JXbObvCAALz2AIcM92aJLd\nk/xea+27rbXvtdb+afbcc5K8pbV2Y2vtziQnJjlmNkwmS8PUSbPXfDbJZ5IcMHvu9Nnrf+iYJH+z\nueFaa9cn+VZVHZjkF5Ocm+SWqnrE7P4nZ4c+Ncn1rbXT2pLPJPm7JL+2jrf9tSTvaq1d1Vr7bpKT\n1nHMWa21S1trdyd5T5ID13p++abz+1n6c/iw1tpdrbV/3NzfEwAWncEQ4J7tIUlunA1Aa/vJJDcu\nu39jknsl2W3ZY7ctu31nkvvNbv9dksOqareqOjzJXa21C7cw4wVJfilLg+D5s58jsrTJ/GFtc6/Z\n+b4++7kjS4Ppbmu/2ez3umnZ/Zvy45XWW5fdXv57rcsbs7RVPK+qrq2ql2/8VwKAabnX2AEA2Kpu\nSrJnVW2zjuHwliwNXD+0V5a2Y7dlaaBcr9baN6rqvCxtCn8q/1FP3RL/kORXkzw0S5XOf0vy60kO\nS/Jny36P81trT9qE91uTZI9l9/fMgA+Uaa19J8l/T/Lfq+qnk3yiqv6ltfaJLX1PAFg0NoYA92z/\nkqVB6Q1VdZ+qundV/fzsudOTvLSqHlpV98vSUHbGsgFygx8cM3v9cVm61vBHaqRVde8k283ubje7\nvz4/3Bhu31q7JUv10Sdn6Zq+f50d8+Ekj6iq51bVvarqP1fVwVW13zre771JfnP2YTX3SfLKjfwe\na7s1S9dc/vB3+ZWqevjs7reS/CBLn7wKAPcYBkOAe7DZkPerWfoAmS9lafP2rNnT/0+Sv87Sxu6L\nWapU/tflL1/77da6f/bsfde01j631nP/X5Jvzl5z1ey915fxmiwNXP8wu/+tWZ4LW2tt9ti3kzwx\nSxvKW2Y/b0jyYwNna+2cJH+a5BNJrk5y0eypf19fhrWclOS0WWX1mbPf8WNV9a0k/5jkz1trPpkU\ngHuUmv07FwDukWafXPq5JPdez7WWANA9G0MA7nGq6ulVtW1V7Zjk5CRnGwoBYP0MhgDcE704yVeS\nXJOlD9R5ybhxAGCxqZICAAB0zsYQAACgc1v9ewyrykoSAABYca21SiYxc9zYWnvo2CE2ZKtXSauq\nHXnkkYPe47rrrsvee++98QM34Ljjjhv0+pV0yy23jB1h7sQTTxw7QpLkj//4j8eOMHfYYYcNfo/T\nTjtt8N9zQ/+eX0mLkuWoo44aO8LcM5/5zLEjzL3qVa8a9Po77rgjO+644+Ac119//eD3WCnPetaz\nNn7QKnj9618/doS5E044YewISZKPfOQjg9/jpJNOykknnTT4fc4///zB77FSzjrrrLEjJFmcHEny\npS99aewISZKnPOUpg15/zTXXZN99912RLAcddNCKvM9Qi/Lv40MOOWRKg+E866JSJQUAAOicwRAA\nAKBzW/0aw5WwEhUnWE0HHHDA2BFgk2233XZjR4DNcsQRR4wdATbZTjvtNHaEblQtblNzCt8EMYmN\nocGQqTEYMiXbb7/92BFgsxgMmZKdd9557AiwSSYxGAIAALD1TKJKCgAAsCGqpMPYGAIAAHTOYAgA\nANA5VVIAAGDyFrlKOgU2hgAAAJ0bNBhW1ZOr6qqqurqqXr5SoQAAAFg9W1wlraptkrw1yeOT3JLk\n4qr6YGvtqpUKBwAAsClUSYcZsjE8NMk1rbUbW2vfT3JGkqNWJhYAAACrZchg+OAkNy27f/PsMQAA\nACbEp5ICAACTp0o6zJDB8MtJ9lx2f4/ZYz/muuuum9/ecccds+OOOw44LQAA0JtLL700l1566dgx\n7rGGDIYXJ9mnqvZKsibJMUmOXdeBe++994DTAAAAvTvooINy0EEHze+fcsopI6a559niwbC1dldV\nHZ/kvCxdq3hqa+3KFUsGAACwiVRJhxl0jWFr7Zwk+61QFgAAAEYw6AvuAQAAmD6fSgoAAEyeKukw\nNoYAAACdMxgCAAB0TpUUAACYPFXSYWwMAQAAOmcwBAAA6JwqKQAAMHmqpMPYGAIAAHTOYAgAANA5\ngyEAAEDnXGMIAABMnmsMh7ExBAAA6NyqbAxf+9rXrsZpNuiSSy4ZO8LcO97xjrEjzJ144oljR0iS\nPPKRjxw7wtx3v/vdsSMkSV7ykpeMHWHunHPOGTtCkuSQQw4ZO8LcLbfcMnaEOf+H9Mc98IEPHDtC\nkuRLX/rS2BHmDj744LEjLJznPe95Y0eY23bbbceOkCQ54ogjxo6wcBbpn/e33nrr2BGSJJdffvnY\nEdgKVEkBAIDJ8z9Kh1ElBQAA6JzBEAAAoHOqpAAAwOSpkg5jYwgAANA5gyEAAEDnVEkBAIDJUyUd\nxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1TJQUAACZPlXQYG0MAAIDOGQwBAAA6\np0oKAABMnirpMDaGAAAAnTMYAgAAdE6VFAAAmDxV0mFsDAEAADpnMAQAAOicKikAADB5qqTD2BgC\nAAB0zmAIAADQOVVSAABg8lRJh7ExBAAA6JzBEAAAoHMGQwAAgM65xhAAAJg81xgOY2MIAADQOYMh\nAABA51RJAQCAyVMlHcbGEAAAoHOrsjG89tprV+M0G3TmmWeOHWFu++23HzvCwnnPe94zdoS5Sy+9\ndOwISZKHPvShY0dYOL/yK78ydoS5173udWNHmDv++OPHjrBwXvWqV40dIUly+umnjx1h7rWvfe3Y\nEZIkf/RHfzR2hLljjz127Ahzz3zmM8eOkCT5wQ9+MHaEhbP77ruPHWHu13/918eOkCS5+uqrx46Q\nJPnABz4wdoTBqurJSf5nlhZ2p7bWTl7PcYck+ackz26tnTl77IYk/5bk7iTfb60dOiSLKikAADB5\nU6uSVtU2Sd6a5PFJbklycVV9sLV21TqOe0OSc9d6i7uTHNFau2Ml8qiSAgAArL5Dk1zTWruxtfb9\nJGckOWodx52Q5P1JvrLW45UVnOcMhgAAAKvvwUluWnb/5tljc1X1k0me3lr7iywNgsu1JB+tqour\n6kVDw6iSAgAAkze1Kukm+p9JXr7s/vJf8rGttTVVtUuWBsQrW2sXbumJDIYAAAAr6Hvf+16+973v\nbeywLyfZc9n9PWaPLXdwkjNqaep9YJL/UlXfb62d3VpbkySttdur6qwsVVMNhgAAAItg2223zbbb\nbju//53vfGddh12cZJ+q2ivJmiTHJPmRj0ture39w9tV9a4kH2qtnV1V90myTWvt21V13yRPTPKa\nIZkNhgAAwORNrUraWrurqo5Pcl7+4+sqrqyqFy893d6x9kuW3d4tyVlV1bI0072ntXbekDwGQwAA\ngBG01s5Jst9aj/3leo59wbLb1yc5cCWz+FRSAACAztkYAgAAkze1KumisTEEAADonMEQAACgc6qk\nAADA5KmSDmNjCAAA0DmDIQAAQOdUSQEAgMlTJR3GxhAAAKBzBkMAAIDOqZICAACTp0o6jI0hAABA\n5wyGAAAAnVMlBQAAJk+VdBgbQwAAgM4ZDAEAADpnMAQAAOicawwBAIDJc43hMDaGAAAAnTMYAgAA\ndE6VFAAAmDxV0mFsDAEAADpnMAQAAOicKikAADB5qqTD2BgCAAB0zmAIAADQuWqtbd0TVLWtfQ4A\nAKAvVZXWWs1ut4c97GFjR1qv66+/fp51UdkYAgAAdM5gCAAA0DmfSgoAAEyeTyUdxsYQAACgcwZD\nAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1TJQUAACZPlXQYG0MAAIDOGQwBAAA6p0oKAABMnirp\nMDaGAAAAnTMYAgAAdE6VFAAAmDxV0mFsDAEAADpnMAQAAOicwRAAAKBzrjEEAAAmzzWGw9gYAgAA\ndM5gCAAA0DlVUgAAYPJUSYexMQQAAOicwRAAAKBzqqQAAMDkqZIOY2MIAADQOYMhAABA51RJAQCA\nyVMlHcbGEAAAoHMGQwAAgM6pkgIAAJOnSjqMjSEAAEDnDIYAAACdW5Uq6Yknnrgap9mg7bfffuwI\ncx/72MfGjjD3D//wD2NHSJL8xm/8xtgR5r73ve+NHSFJ8rjHPW7sCHPHH3/82BGSJK985SvHjjD3\nz//8z2NHmHv9618/doQkycEHHzx2hLlHP/rRY0dIkjzqUY8aO8LC+eu//uuxI8wdffTRY0eY22GH\nHcaOkCR54QtfOHaEuV/4hV8YO0KS5Atf+MLYEebe+MY3jh0hSXLRRReNHWGdVEmHsTEEAADonMEQ\nAACgcz6VFAAAmDxV0mFsDAEAADpnMAQAAOicKikAADB5qqTD2BgCAAB0zmAIAADQOVVSAABg8lRJ\nh7ExBAAA6JzBEAAAoHOqpAAAwOSpkg5jYwgAANA5gyEAAEDnDIYAAACdc40hAAAwea4xHMbGEAAA\noHMGQwAAgM6pkgIAAJOnSjqMjSEAAEDnDIYAAACdUyUFAAAmT5V0GBtDAACAzhkMAQAARlBVT66q\nq6rq6qp6+Tqef1pVfaaq/rWqLqmqIzf1tZtLlRQAAJi8qVVJq2qbJG9N8vgktyS5uKo+2Fq7atlh\nH2utnT07fv8kZyXZZxNfu1lsDAEAAFbfoUmuaa3d2Fr7fpIzkhy1/IDW2p3L7t4vyVc39bWby2AI\nAACw+h6c5KZl92+ePfYjqurpVXVlko8k+a+b89rNoUoKAABM3tSqpJuqtfaBJB+oqscl+esk+22N\n8xgMAQAAVtA3v/nNfOtb39rYYV9Osuey+3vMHlun1tonq+peVbXz5r52UxgMAQAAVtADHvCAPOAB\nD5jfv+WWW9Z12MVZ+iCZvZKsSXJMkmOXH1BVD2+tfXF2+9FJ0lr7WlV9Y2Ov3VwGQwAAYPKmViVt\nrd1VVccnOS9Ln/1yamvtyqp68dLT7R1JnlFVxyX5XpLvZGkAXO9rh+QxGAIAAIygtXZO1rpmsLX2\nl8tuvzHJGzf1tUP4VFIAAIDO2RgCAACTN7Uq6aJZlcHw6quvXo3TbNBNN9208YNWyW233TZ2hIXz\nB3/wB2NHmNtmm8VYpJ955pljR1g4i/LXJkn22GOPsSPM3XrrrWNHWDhHH3302BGSJJ/85CfHjjC3\n8847jx1h4XzjG98YO8LcddddN3aEJMnv/u7vjh1h7qKLLho7QpLk8ssvHzvC3OMe97ixIyyUK68c\ndEkda1mc/8oCAABgFKqkAADA5KmSDmNjCAAA0DmDIQAAQOdUSQEAgMlTJR3GxhAAAKBzBkMAAIDO\nqZICAACTp0o6jI0hAABA5wyGAAAAnTMYAgAAdM41hgAAwOS5xnCYQYNhVd2Q5N+S3J3k+621Q1ci\nFAAAAKtn6Mbw7iRHtNbuWIkwAAAArL6hg2HFdYoAAMDIVEmHGTrUtSQfraqLq+pFKxEIAACA1TV0\nY/jY1tqaqtolSwPila21C1ciGAAAAKtj0GDYWlsz++PtVXVWkkOT/NhgeMUVV8xv77LLLtlll12G\nnBYAAOjMmjVrsmbNmvU+r0o6zBYPhlV1nyTbtNa+XVX3TfLEJK9Z17E//dM/vaWnAQAAyO67757d\nd999fv/Tn/70iGnueYZsDHdLclZVtdn7vKe1dt7KxAIAAGC1bPFg2Fq7PsmBK5gFAABgi6iSDuOr\nJgAAADpnMAQAAOjc0K+rAAAAGJ0q6TA2hgAAAJ0zGAIAAHROlRQAAJg8VdJhbAwBAAA6ZzAEAADo\nnCopAAAweaqkw9gYAgAAdM5gCAAA0DlVUgAAYPJUSYdZlcHwW9/61mqcZoP+9m//duwIc1/84hfH\njrBwnv3sZ48dYe7e97732BGSJC9/+cvHjrBwzjzzzLEjzD3qUY8aO8Lc+9///rEjJEme+tSnjh1h\nbr/99hs7QpLkyCOPHDvC3OMf//ixIyyct73tbWNHmPvMZz4zdoQkyQc+8IGxIyycRzziEWNHmHvv\ne987doQkydVXXz12BLYCVVIAAIDOqZICAACTp0o6jI0hAABA5wyGAAAAnVMlBQAAJk+VdBgbQwAA\ngM4ZDAEAADpnMAQAAOicawwBAIDJc43hMDaGAAAAnTMYAgAAdE6VFAAAmDxV0mFsDAEAADpnMAQA\nAOicKikAADB5qqTD2BgCAAB0zmAIAADQOVVSAABg8lRJh7ExBAAA6JzBEAAAoHOqpAAAwOSpkg5j\nYwgAANA5gyEAAEDnVEkBAIDJUyUdxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1T\nJQUAACZPlXQYG0MAAIDOGQwBAAA6p0oKAABMnirpMDaGAAAAnavW2tY9QVXb2ucAAAD6UlVprdXs\ndvvVX/3VsSOt14c+9KF51kWlSgoAAEyeKukwqqQAAACdMxgCAAB0zmAIAADQOdcYAgAAk+caw2Fs\nDAEAADpnMAQAAOicKikAADB5qqTD2BgCAAB0zmAIAADQOVVSAABg8lRJh7ExBAAA6JzBEAAAoHMG\nQwAAYPKqamF/NpD5yVV1VVVdXVUvX8fz+1XVP1XVd6vqd9Z67oaq+kxV/WtV/cvQP3+uMQQAAFhl\nVbVNkrcmeXySW5JcXFUfbK1dteywryU5IcnT1/EWdyc5orV2x0rksTEEAABYfYcmuaa1dmNr7ftJ\nzkhy1PIDWmtfba1dmuQH63h9ZQXnORtDAABg8ib4qaQPTnLTsvs3Z2lY3FQtyUer6q4k72itnTIk\njMEQAABgeh7bWltTVbtkaUC8srV24Za+mcEQAABgBd1+++25/fbbN3bYl5Psuez+HrPHNklrbc3s\nj7dX1VlZ2jYaDAEAgH4tUpV01113za677jq/f+WVV67rsIuT7FNVeyVZk+SYJMdu4G3nv2BV3SfJ\nNq21b1fVfZM8MclrhmQ2GAIAAKyy1tpdVXV8kvOy9CEyp7bWrqyqFy893d5RVbsluSTJ/ZPcXVX/\nLclPJ9klyVlV1bI0072ntXbekDwGQwAAgBG01s5Jst9aj/3lstu3JXnIOl767SQHrmQWgyEAADB5\ni1QlnSLfYwgAANA5gyEAAEDnVEkBAIDJUyUdxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcM\nhgAAAJ1TJQUAACZPlXQYG0MAAIDOGQwBAAA6102V9Pzzzx87wty55547doS517/+9WNHSJJcdNFF\nY0eY+/M///OxIyRJLrzwwrEjzN1www1jR0iS7LDDDmNHmHvzm988doS5j3zkI2NHSJKceeaZY0eY\nO+qoo8aOkCR5wQteMHaEufe9731jR0iSvPvd7x47wtzJJ588doS5F73oRWNHSJKcddZZY0eYe+EL\nXzh2hCTJJZdcMnaEuRNPPHHsCEmSa6+9duwIbAXdDIYAAMA9l2sMh1ElBQAA6JzBEAAAoHOqpAAA\nwOSpkg5jYwgAANA5gyEAAEDnVEkBAIDJUyUdxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcM\nhgAAAJ1TJQUAACZPlXQYG0MAAIDOGQwBAAA6p0oKAABMnirpMDaGAAAAnTMYAgAAdE6VFAAAmDxV\n0mFsDAEAADpnMAQAAOicKikAADB5qqTD2BgCAAB0zmAIAADQOVVSAABg8lRJh7ExBAAA6JzBEAAA\noHOqpAAAwOSpkg5jYwgAANA5gyEAAEDnDIYAAACdc40hAAAwea4xHMbGEAAAoHOrsjE8+uijV+M0\nG3TXXXeNHWHuvve979gRmIBnPetZY0dYOA9/+MPHjjD3xS9+cewIc3/2Z382doSFs/POO48dYeG8\n5CUvGTvCwnn7298+doS5G2+8cewISZLnPOc5Y0dYOF/96lfHjjC3Zs2asSMkSQ455JCxIyRJbrjh\nhrEj3KOokgIAAJOnSjqMKikAAEDnDIYAAACdUyUFAAAmT5V0GBtDAACAzhkMAQAAOqdKCgAATJ4q\n6TA2hgAAAJ0zGAIAAHROlRQAAJg8VdJhbAwBAAA6ZzAEAADonCopAAAweaqkw9gYAgAAdM5gCAAA\n0DlVUgAAYPJUSYexMQQAAOicwRAAAKBzqqQAAMDkqZIOY2MIAADQOYMhAABA51RJAQCAyVMlHcbG\nEAAAoHMGQwAAgM6pkgIAAJOnSjqMjSEAAEDnDIYAAACdMxgCAAB0zjWGAADA5LnGcBgbQwAAgM4Z\nDAEAADqnSgoAAEyeKukwNoYAAACdMxgCAAB0blWqpJdddtlqnGaD7n//+48dYe7Zz3722BEWzu/8\nzu+MHWHu2GOPHTtCkuSaa64ZO8LCedKTnjR2hLn73ve+Y0eY+/a3vz12hIXz+c9/fuwISZIrrrhi\n7AhzO+6449gRkiR///d/P3aEue22227sCHP77rvv2BGSJHfffffYERbOK17xirEjzB166KFjR0iS\nHHPMMWNHSJK8733v+5H7qqTD2BgCAAB0zmAIAADQOYMhAAAweVW1sD8byPzkqrqqqq6uqpev55g/\nraprqurTVXXg5rx2cxgMAQAAVllVbZPkrUmelORRSY6tqkeudcx/SfLw1tq+SV6c5O2b+trNZTAE\nAABYfYcmuaa1dmNr7ftJzkhy1FrHHJXktCRprf1zkh2qardNfO1m8QX3AADA5E3wU0kfnOSmZfdv\nztLAt7FjHryJr90sNoYAAADTsNWmXxtDAACAFXTTTTfl5ptv3thhX06y57L7e8weW/uYh6zjmG03\n4bWbxWAIAABM3iJVSffcc8/sued/zG2f+tSn1nXYxUn2qaq9kqxJckySY9c65uwkv53kb6vqsCTf\naK3dVlVf3YTXbhaDIQAAwCprrd1VVccnOS9Ll/id2lq7sqpevPR0e0dr7SNV9ZSqujbJd5L85oZe\nOySPwRAAAGAErbVzkuy31mN/udb94zf1tUMYDAEAgMlbpCrpFG30U0mr6tSquq2qPrvssR2r6ryq\n+kJVnVtVO2zdmAAAAGwtm/J1Fe9K8qS1Hvv9JB9rre2X5ONJTlzpYAAAAKyOjVZJW2sXzj7tZrmj\nkhw+u/1XSc7P0rAIAACw6lRJh9nSL7jftbV2W5K01m5NsuvKRQIAAGA1rdSHz7QNPfmNb3xjfnu7\n7bbLdtttt0KnBQAAevCZz3wmn/3sZzd+IFtkSwfD26pqt9mXKz4oyVc2dPBP/MRPbOFpAAAAkgMO\nOCAHHHDA/P673/3uH3lelXSYTa2S1uznh85O8vzZ7ecl+eAKZgIAAGAVbcrXVfxNkn9K8oiq+lJV\n/WaSNyR5QlV9IcnjZ/cBAACYoE35VNLnrOepX17hLAAAAFtElXSYLf1UUgAAAO4hDIYAAACdMxgC\nAAB0bqW+xxAAAGA0rjEcxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1TJQUAACZP\nlXQYG0MAAIDOGQwBAAA6V621rXuCqra1zwEAAPSlqtJaq9nt9opXvGLsSOv1ute9bp51UdkYAgAA\ndM5gCAAA0DmfSgoAAEyeTyUdxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1TJQUA\nACZPlXQYG0MAAIDOGQwBAAA6p0oKAABMnirpMDaGAAAAnTMYAgAAdE6VFAAAmDxV0mFsDAEAADpn\nMAQAAOicKikAADB5qqTD2BgCAAB0zmAIAADQOYMhAABA51xjCAAATJ5rDIexMQQAAOicwRAAAKBz\nqqQAAMDkqZIOY2MIAADQOYMhAABA51RJAQCAyVMlHcbGEAAAoHMGQwAAgM6pkgIAAJOnSjqMjSEA\nAEDnDIYAAACdUyUFAAAmT5V0mFUZDJ/2tKetxmk2aN999x07wtz97ne/sSPMveY1rxk7QpLkZ37m\nZ8aOMHfCCSeMHSFJcvXVV48dYe7Nb37z2BGSJH/yJ38ydoS5gw46aOwIc/e612L8P75f/MVfHDvC\n3Ate8IKxIyRJLrnkkrEjzP3gBz8YO0KS5Iorrhg7wtzLXvaysSPMPexhDxs7QpJk9913HzvC3DOe\n8YyxIyRJjjzyyLEjzC3KP++5Z1IlBQAA6Jz/7QAAAEyeKukwNoYAAACdMxgCAAB0TpUUAACYPFXS\nYWwMAQAAOmcwBAAA6JwqKQAAMHmqpMPYGAIAAHTOYAgAANA5VVIAAGDyVEmHsTEEAADonMEQAACg\nc6qkAADA5KmSDmNjCAAA0DmDIQAAQOcMhgAAAJ1zjSEAADB5rjEcxsYQAACgcwZDAACAzqmSAgAA\nk6dKOozPMWhqAAAZL0lEQVSNIQAAQOcMhgAAAJ1TJQUAACZPlXQYG0MAAIDOGQwBAAA6ZzAEAAAm\nr6oW9mcLfpcdq+q8qvpCVZ1bVTus57hTq+q2qvrsWo+/uqpurqrLZj9P3tg5DYYAAACL5feTfKy1\ntl+Sjyc5cT3HvSvJk9bz3Ftaa4+e/ZyzsRMaDAEAABbLUUn+anb7r5I8fV0HtdYuTHLHet5js1aV\nPpUUAACYvHvYp5Lu2lq7LUlaa7dW1a5b8B7HV9VvJLkkyctaa/+2oYNtDAEAAFZZVX20qj677Odz\nsz8+bR2Ht818+7cl2bu1dmCSW5O8ZWMvsDEEAABYQddee22uvfbaDR7TWnvC+p6bfaDMbq2126rq\nQUm+sjnnb63dvuzuKUk+tLHXGAwBAIDJW6Qq6b777pt99913fv/cc8/d3Lc4O8nzk5yc5HlJPriB\nYytrXU9YVQ9qrd06u3t0kss3dsJVGQz32Wef1TjNBl1++Ub/XKyan/3Znx07wsJ50YteNHaEuU9+\n8pNjR0iSXHPNNWNHWDg/93M/N3aEude85jVjR5j77d/+7bEjLJzHPe5xY0dIslj/kXLAAQeMHWHh\nPPKRjxw7wtxOO+00doQkyac+9amxI8w94xnPGDtCkuQP//APx44w93u/93tjR0iSvPSlLx07QpLk\nox/96NgRtqaTk7y3ql6Q5MYkz0qSqto9ySmttafO7v9NkiOS7FxVX0ry6tbau5K8saoOTHJ3khuS\nvHhjJ7QxBAAAWCCtta8n+eV1PL4myVOX3X/Oel5/3Oae02AIAABM3iK1NKbIp5ICAAB0zmAIAADQ\nOVVSAABg8lRJh7ExBAAA6JzBEAAAoHOqpAAAwOSpkg5jYwgAANA5gyEAAEDnVEkBAIDJUyUdxsYQ\nAACgcwZDAACAzhkMAQAAOucaQwAAYPJcYziMjSEAAEDnDIYAAACdUyUFAAAmT5V0GBtDAACAzhkM\nAQAAOqdKCgAATJ4q6TA2hgAAAJ0zGAIAAHROlRQAAJg8VdJhbAwBAAA6ZzAEAADonCopAAAweaqk\nw9gYAgAAdM5gCAAA0DlVUgAAYPJUSYexMQQAAOicwRAAAKBzqqQAAMDkqZIOY2MIAADQuVXZGN5x\nxx2rcZoNaq2NHWHuEY94xNgRFs4ZZ5wxdoS53XfffewISZKXvvSlY0dYOFddddXYEeZ22mmnsSPM\nvelNbxo7QpLk6U9/+tgR5vbaa6+xIyRJHvOYx4wdYe7rX//62BEWztOe9rSxI8wddthhY0dIkuyx\nxx5jR1g4D3nIQ8aOMHfEEUeMHSFJ8hd/8RdjR2ArUCUFAAAmT5V0GFVSAACAzhkMAQAAOqdKCgAA\nTJ4q6TA2hgAAAJ0zGAIAAHROlRQAAJg8VdJhbAwBAAA6ZzAEAADonMEQAACgc64xBAAAJs81hsPY\nGAIAAHTOYAgAANA5VVIAAGDyVEmHsTEEAADonMEQAACgc6qkAADA5KmSDmNjCAAA0DmDIQAAQOdU\nSQEAgMlTJR3GxhAAAKBzBkMAAIDOqZICAACTp0o6jI0hAABA5wyGAAAAnVMlBQAAJk+VdBgbQwAA\ngM4ZDAEAADqnSgoAAEyeKukwNoYAAACdMxgCAAB0TpUUAACYPFXSYWwMAQAAOmcwBAAA6Fy11rbu\nCara1j4HAADQl6pKa61mt9vpp58+dqT1OvbYY+dZF5WNIQAAQOcMhgAAAJ3zqaQAAMDk+VTSYWwM\nAQAAOmcwBAAA6JzBEAAAoHOuMQQAACbPNYbD2BgCAAB0zmAIAADQOVVSAABg8lRJh7ExBAAA6NxG\nB8OqOrWqbquqzy577NVVdXNVXTb7efLWjQkAAMDWsilV0ncl+bMkp631+Ftaa29Z+UgAAACbR5V0\nmI1uDFtrFya5Yx1P+TMPAABwDzDkGsPjq+rTVfXOqtphxRIBAACwqrZ0MHxbkr1bawcmuTWJSikA\nADCaqlrYny34XXasqvOq6gtVde66FnFVde+q+ueq+teq+nxVvW5zXr+2Lfq6itba7cvunpLkQxs6\n/qSTTprfPuKII3LEEUdsyWkBAIBOnX/++Tn//PPHjrFafj/Jx1prb6yqlyc5cfbYXGvt36vql1pr\nd1bVf0ryj1X12NbaP27K69dWrbWNpqqqhyb5UGtt/9n9B7XWbp3dfmmSQ1prz1nPa9umnAMAAGBT\nVVVaazW73d7//vePHWm9nvnMZ86zboqquirJ4a2126rqQUnOb609cgPH3yfJ+Ume31q7YnNfn2zC\nxrCq/ibJEUl2rqovJXl1kl+qqgOT3J3khiQv3pRfEAAAYGu4h30q6a6ttduSpLV2a1Xtuq6Dqmqb\nJJcmeXiSt7fWrtic1y+30cFwPZvAd23sdQAAAKxbVX00yW7LH0rSkrxyHYevs4LZWrs7yc9W1QOS\nnFdVh7fWLtjU1y+3RdcYAgAAsG6XX355Pv/5z2/wmNbaE9b3XFXdVlW7LauCfmUj7/XNqvrfSQ5O\nckGSzXp9YjAEAADuARapSrr//vtn//33n99/73vfu7lvcXaS5yc5Ocnzknxw7QOq6oFJvt9a+7eq\n2j7JE5K8ZlNfv7Yh32MIAADAyjs5yROq6gtJHp/kDUlSVbtX1Ydnx+ye5BNV9a9JPpXk7Nba/7uh\n12+IjSEAAMACaa19Pckvr+PxNUmeOrv9uSSP3pzXb4jBEAAAmLxFqpJOkSopAABA5wyGAAAAnVMl\nBQAAJk+VdBgbQwAAgM4ZDAEAADqnSgoAAEyeKukwNoYAAACdMxgCAAB0blWqpCeccMJqnGaDdthh\nh7EjzF1yySVjR5g755xzxo6QJDn88MPHjjB35513jh0hSfK1r31t7Ahz11133dgRkiSnnHLK2BHm\nDjvssLEjzJ1++uljR0iSvO51rxs7wtw73/nOsSMkSS677LKxI8z9/M///NgRkiTPfe5zx44wt8su\nu4wdYe6nfuqnxo6QJDnmmGPGjjD3kpe8ZOwISZLTTjtt7AhzP/mTPzl2hCSL9d/Vy6mSDmNjCAAA\n0DmDIQAAQOcMhgAAAJ3zdRUAAMDkucZwGBtDAACAzhkMAQAAOqdKCgAATJ4q6TA2hgAAAJ0zGAIA\nAHROlRQAAJg8VdJhbAwBAAA6ZzAEAADonCopAAAweaqkw9gYAgAAdM5gCAAA0DlVUgAAYPJUSYex\nMQQAAOicwRAAAKBzqqQAAMDkqZIOY2MIAADQOYMhAABA51RJAQCAyVMlHcbGEAAAoHMGQwAAgM6p\nkgIAAJOnSjqMjSEAAEDnDIYAAACdUyUFAAAmT5V0GBtDAACAzhkMAQAAOqdKCgAATJ4q6TA2hgAA\nAJ0zGAIAAHTOYAgAANC5VbnGcJdddlmN02zQQx/60LEjzD33uc8dO8LCedOb3jR2hLk777xz7AhJ\nkk984hNjR1g4b3vb28aOMPfOd75z7Ahz97nPfcaOsHAW5Z/5O+2009gR5j7+8Y+PHSHJYv07cJ99\n9hk7wtwTn/jEsSMkSb7yla+MHWHh3P/+9x87wtyi/PU59dRTx46wTq4xHMbGEAAAoHMGQwAAgM75\nugoAAGDyVEmHsTEEAADonMEQAACgc6qkAADA5KmSDmNjCAAA0DmDIQAAQOdUSQEAgMlTJR3GxhAA\nAKBzBkMAAIDOqZICAACTp0o6jI0hAABA5wyGAAAAnVMlBQAAJk+VdBgbQwAAgM4ZDAEAADqnSgoA\nAEyeKukwNoYAAACdMxgCAAB0TpUUAACYPFXSYWwMAQAAOmcwBAAA6JwqKQAAMHmqpMPYGAIAAHTO\nYAgAANA5VVIAAGDyVEmHsTEEAADonMEQAACgcwZDAACAzrnGEAAAmDzXGA5jYwgAANA5gyEAAEDn\nVqVKuv/++6/GaTZo5513HjvC3IEHHjh2hLnvfve7Y0dIknz84x8fO8LcBRdcMHaEJMmTnvSksSMs\nnF133XXsCHOPecxjxo4w96lPfWrsCAtnUf6cXHbZZWNHmPut3/qtsSMsnBNOOGHsCHOHH3742BGS\nJDvuuOPYERbO6aefPnaEuW9+85tjR0iSXHjhhWNHWCdV0mFsDAEAADpnMAQAAOicTyUFAAAmT5V0\nGBtDAACAzhkMAQAAOmcwBAAAJq+qFvZnC36XHavqvKr6QlWdW1U7bODYbarqsqo6e9ljr66qm2eP\nX1ZVT97YOQ2GAAAAi+X3k3ystbZfko8nOXEDx/63JFes4/G3tNYePfs5Z2MnNBgCAAAslqOS/NXs\n9l8lefq6DqqqPZI8Jck71/X05pzQYAgAAEze2HXRlaySJtm1tXZbkrTWbk2y63qO+7+T/G6Sto7n\njq+qT1fVOzdURf0hgyEAAMAqq6qPVtVnl/18bvbHp63j8B8b/KrqV5Lc1lr7dJa2g8sn0Lcl2bu1\ndmCSW5O8ZWN5fI8hAADACrr00ktz6aWXbvCY1toT1vdcVd1WVbu11m6rqgcl+co6DntskqdV1VOS\nbJ/k/lV1WmvtuNba7cuOOyXJhzaW2WAIAABM3iJ9wf3BBx+cgw8+eH7/lFNO2dy3ODvJ85OcnOR5\nST649gGttVckeUWSVNXhSV7WWjtudv9Bswpqkhyd5PKNnVCVFAAAYLGcnOQJVfWFJI9P8oYkqard\nq+rDm/D6N85qqZ9OcniSl27sBTaGAAAAC6S19vUkv7yOx9ckeeo6Hr8gyQXL7h+3uec0GAIAAJO3\nSFXSKVIlBQAA6JzBEAAAoHOqpAAAwOSpkg5jYwgAANA5gyEAAEDnVEkBAIDJUyUdxsYQAACgcwZD\nAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ0zGAIAAHTONYYAAMDkucZwGBtDAACAzhkMAQAAOqdK\nCgAATJ4q6TA2hgAAAJ0zGAIAAHROlRQAAJg8VdJhqrW2dU9Q1bb2OQAAgL5UVVprNbvdPve5z40d\nab3233//edZFpUoKAADQOVVSAABg8lRJh7ExBAAA6JzBEAAAoHOqpAAAwOSpkg5jYwgAANA5gyEA\nAEDnVEkBAIDJUyUdxsYQAACgcwZDAACAzqmSAgAAk6dKOoyNIQAAQOcMhgAAAJ1TJQUAACZPlXQY\nG0MAAIDOGQwBAAA6p0oKAABMnirpMDaGAAAAnTMYAgAAdE6VFAAAmDxV0mFsDAEAADpnMAQAAOic\nwRAAAKBzrjEEAAAmzzWGw9gYAgAAdM5gCAAA0DlVUgAAYPJUSYexMQQAAOicwRAAAKBzqqQAAMDk\nqZIOY2MIAADQOYMhAABA51RJAQCAyVMlHWZVBsMPf/jDq3GaDXrta187doS5T3/602NHmPv3f//3\nsSMkyf/f3h2FaFZYdwD/n1YK1tBFhV1Lt40ppSbIWl1IStwWtjSpIQSUPoTWPCiF0geLgbysLgXx\npUSQUKX0oW0iS4iURKgrpVQjy1K6sCQhcbVqTREVLdkxpS3iWyWnD/M5md3OzM74OXPv9f5+8LH3\n3rl3vjNw9/L955x7Jw8//PDQJaw5duzY0CUkSR5//PGhS1hzyy23DF0CAAC7yCgpAADAzBklBQAA\nJs8o6XJ0DAEAAGZOMAQAAJg5o6QAAMDkGSVdjo4hAADAzAmGAAAAM2eUFAAAmDyjpMvRMQQAAJg5\nwRAAAGDmjJICAACTZ5R0OTqGAAAAM3fJYFhVB6vqVFU9X1XPVdXdi+1XVtVTVfVSVT1ZVft2v1wA\nAADeb9vpGL6T5EvdfX2STya5q6o+muSeJE9393VJTiW5d/fKBAAA2FxVjfY1BZcMht19vrufWSy/\nneTFJAeT3JrkxGK3E0lu260iAQAA2D07usewqq5NcmOSs0kOdPdKshoek+x/v4sDAABg9237qaRV\n9aEkjyX5Yne/XVV90S4XrwMAAOyJqYxsjtW2gmFVXZbVUPj17j652LxSVQe6e6Wqrkny5mbHP/ro\no2vLhw4dyqFDh5YoGQAAmJvTp0/n9OnTQ5fxgbXdjuHXkrzQ3Q+t2/ZEkjuTPJDkjiQnNzguSXL7\n7be/1/oAAABy9OjRHD16dG39/vvvH66YD6BLBsOqOpLkC0meq6ofZHVk9HhWA+E3q+qPkryW5PO7\nWSgAAAC745LBsLvPJPnZTb78qfe3HAAAgJ1zj+FydvRUUgAAAD54BEMAAICZ2/afqwAAABgro6TL\n0TEEAACYOcEQAABg5oySAgAAk2eUdDk6hgAAADMnGAIAAMycUVIAAGDyjJIuR8cQAABgRKrqyqp6\nqqpeqqonq2rfJvvtq6pvVdWLVfV8Vf3mTo5fTzAEAAAYl3uSPN3d1yU5leTeTfZ7KMk/dvfHkvxG\nkhd3ePwawRAAAJi8qhrt6z24NcmJxfKJJLdt8PP+QpLf7u5HkqS73+nut7Z7/MUEQwAAgHHZ390r\nSdLd55Ps32CfjyT5z6p6pKq+X1V/XVWX7+D4C+zJw2euvvrqvXibLV111VVDl7Dm+PHjQ5cwOmfP\nnh26hDX79l1yBHtP3HzzzUOXAADALqmqbyc5sH5Tkk7yZxvs3htsuyzJ4SR3dff3quovsjpCet/i\ne13q+P/3zQAAACZtTE8lPXPmTM6cObPlPt396c2+VlUrVXWgu1eq6pokb26w2xtJXu/u7y3WH0ty\nbLF8fhvHX0AwBAAAeB8dOXIkR44cWVt/8MEHd/otnkhyZ5IHktyR5OTFOyxC3+tV9evd/cMkv5vk\nhe0efzH3GAIAAIzLA0k+XVUvZTXwfTlJquoXq+of1u13d5JvVNUzWX0q6Z9vdfxWdAwBAIDJG9Mo\n6bK6+7+SfGqD7T9K8rl16+eSfHy7x29FxxAAAGDmBEMAAICZM0oKAABM3gdplHQIOoYAAAAzJxgC\nAADMnFFSAABg8oySLkfHEAAAYOYEQwAAgJkzSgoAAEyeUdLl6BgCAADMnGAIAAAwc4IhAADAzLnH\nEAAAmDz3GC5HxxAAAGDmBEMAAICZM0oKAABMnlHS5egYAgAAzJxgCAAAMHNGSQEAgMkzSrocHUMA\nAICZEwwBAABmzigpAAAweUZJl6NjCAAAMHOCIQAAwMwZJQUAACbPKOlydAwBAABmTjAEAACYOaOk\nAADA5BklXY6OIQAAwMwJhgAAADO3J6Okb7311l68zZYuv/zyoUtYc8UVVwxdwujcdNNNQ5ew5vDh\nw0OXkCR55ZVXhi5hzQ033DB0CQAAWzJKuhwdQwAAgJkTDAEAAGbOU0kBAIDJM0q6HB1DAACAmRMM\nAQAAZs4oKQAAMHlGSZejYwgAADBzgiEAAMDMGSUFAAAmzyjpcnQMAQAAZk4wBAAAmDnBEAAAYObc\nYwgAAEyeewyXo2MIAAAwc4IhAADAzBklBQAAJs8o6XJ0DAEAAGZOMAQAAJg5o6QAAMDkGSVdjo4h\nAADAzAmGAAAAM2eUFAAAmDyjpMvRMQQAAJg5wRAAAGDmjJICAACTZ5R0OTqGAAAAMycYAgAAzJxR\nUgAAYPKMki5HxxAAAGDmJhEMz507N3QJsCMvv/zy0CXAtp0+fXroEmBHnLNMifOVqZhEMHz22WeH\nLgF2RDBkSnxoYWqcs0yJ83XvVNVoX1MwiWAIAADA7hEMAQAAZq66e3ffoGp33wAAAJil7q4kqapX\nk3x42Gq29Fp3Xzt0EVvZ9WAIAADAuBklBQAAmDnBEAAAYOZGHwyr6jNV9W9V9cOqOjZ0PbCVqnq1\nqs5V1Q+q6jtD1wMXq6qvVtVKVT27btuVVfVUVb1UVU9W1b4ha4T1Njln76uqN6rq+4vXZ4asEd5V\nVQer6lRVPV9Vz1XV3YvtrrOM3qiDYVX9TJK/THJLkuuT/GFVfXTYqmBLP0lytLtv6u5PDF0MbOCR\nrF5T17snydPdfV2SU0nu3fOqYHMbnbNJ8pXuPrx4/dNeFwWbeCfJl7r7+iSfTHLX4rOr6yyjN+pg\nmOQTSf69u1/r7v9N8ndJbh24JthKZfz/r5ix7v6XJP990eZbk5xYLJ9IctueFgVb2OScTVavtzAq\n3X2+u59ZLL+d5MUkB+M6ywSM/QPsLyV5fd36G4ttMFad5NtV9d2q+uOhi4Ft2t/dK8nqh5ok+weu\nB7bjT6vqmar6W2N5jFFVXZvkxiRnkxxwnWXsxh4MYWqOdPfhJJ/N6vjIbw1dELwH/o4RY/dXSX61\nu29Mcj7JVwauBy5QVR9K8liSLy46hxdfV11nGZ2xB8P/SPIr69YPLrbBKHX3jxb//jjJ32d1HBrG\nbqWqDiRJVV2T5M2B64EtdfeP+6d/iPlvknx8yHpgvaq6LKuh8OvdfXKx2XWW0Rt7MPxukl+rqg9X\n1c8l+YMkTwxcE2yoqn5+8RvCVNUVSX4vyb8OWxVsqHLh/VlPJLlzsXxHkpMXHwADu+CcXXywftfv\nx7WWcflakhe6+6F121xnGb366S/cxmnxCOqHshpiv9rdXx64JNhQVX0kq13CTnJZkm84Xxmbqno0\nydEkVydZSXJfkseTfCvJLyd5Lcnnu/t/hqoR1tvknP2drN679ZMkryb5k3fv34IhVdWRJP+c5Lms\nfh7oJMeTfCfJN+M6y4iNPhgCAACwu8Y+SgoAAMAuEwwBAABmTjAEAACYOcEQAABg5gRDAACAmRMM\nAQAAZk4wBAAAmDnBEAAAYOb+DySKOX+LjtIqAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6105a39450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize weights\n",
    "W = model.layers[0].W.get_value(borrow=True)\n",
    "W = np.squeeze(W)\n",
    "print(\"W shape : \", W.shape)\n",
    "\n",
    "pl.figure(figsize=(15, 15))\n",
    "pl.title('conv1 weights')\n",
    "nice_imshow(pl.gca(), make_mosaic(W, 6, 6), cmap=cm.binary)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C1 shape :  (32, 26, 26)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA4QAAAOiCAYAAAAouxkQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n9N4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6105a39690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize convolution result (after activation)\n",
    "C1 = convout1_f(X)\n",
    "C1 = np.squeeze(C1)\n",
    "print(\"C1 shape : \", C1.shape)\n",
    "\n",
    "pl.figure(figsize=(15, 15))\n",
    "pl.suptitle('convout1')\n",
    "nice_imshow(pl.gca(), make_mosaic(C1, 6, 6), cmap=cm.binary)"
   ]
  },
  {
   "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
}