Conv2D_testing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import six\n",
    "import sys\n",
    "import os\n",
    "from optparse import OptionParser\n",
    "from keras.models import load_model, Model, model_from_json\n",
    "from argparse import ArgumentParser\n",
    "from keras import backend as K\n",
    "import numpy as np\n",
    "# fix random seed for reproducibility\n",
    "seed = 42\n",
    "np.random.seed(seed)\n",
    "import h5py\n",
    "import matplotlib\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import preprocessing\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n",
    "sys.path.append(os.getcwd() + '/..')\n",
    "from models.constraints import ZeroSomeWeights\n",
    "from keras.utils.generic_utils import get_custom_objects\n",
    "get_custom_objects().update({\"ZeroSomeWeights\": ZeroSomeWeights})\n",
    "import yaml\n",
    "from train import parse_config, get_features\n",
    "from eval import makeRoc, plot_confusion_matrix\n",
    "from layers.quantized_layers import Clip, BinaryDense, TernaryDense, QuantizedDense\n",
    "from models.models import binary_tanh, ternary_tanh, quantized_relu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('j_ptfrac', 'j_pt', 'j_eta', 'j_mass', 'j_tau1_b1', 'j_tau2_b1', 'j_tau3_b1', 'j_tau1_b2', 'j_tau2_b2', 'j_tau3_b2', 'j_tau32_b1', 'j_tau32_b2', 'j_zlogz', 'j_c1_b0', 'j_c1_b1', 'j_c1_b2', 'j_c2_b1', 'j_c2_b2', 'j_d2_b1', 'j_d2_b2', 'j_d2_a1_b1', 'j_d2_a1_b2', 'j_m2_b1', 'j_m2_b2', 'j_n2_b1', 'j_n2_b2', 'j_tau1_b1_mmdt', 'j_tau2_b1_mmdt', 'j_tau3_b1_mmdt', 'j_tau1_b2_mmdt', 'j_tau2_b2_mmdt', 'j_tau3_b2_mmdt', 'j_tau32_b1_mmdt', 'j_tau32_b2_mmdt', 'j_c1_b0_mmdt', 'j_c1_b1_mmdt', 'j_c1_b2_mmdt', 'j_c2_b1_mmdt', 'j_c2_b2_mmdt', 'j_d2_b1_mmdt', 'j_d2_b2_mmdt', 'j_d2_a1_b1_mmdt', 'j_d2_a1_b2_mmdt', 'j_m2_b1_mmdt', 'j_m2_b2_mmdt', 'j_n2_b1_mmdt', 'j_n2_b2_mmdt', 'j_mass_trim', 'j_mass_mmdt', 'j_mass_prun', 'j_mass_sdb2', 'j_mass_sdm1', 'j_multiplicity', 'j_g', 'j_q', 'j_w', 'j_z', 'j_t', 'j_undef')\n"
     ]
    }
   ],
   "source": [
    "data_path = '../data/processed-pythia82-lhc13-all-pt1-50k-r1_h022_e0175_t220_nonu_truth.z'\n",
    "import h5py\n",
    "f = h5py.File(data_path, 'r')\n",
    "treeArray = f['t_allpar_new'][()]\n",
    "print(treeArray.dtype.names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading configuration from train_config_conv2d.yml\n",
      "{'Inputs': ['j1_etarot', 'j1_phirot', 'j1_ptrel', 'j_index'], 'Labels': ['j_g', 'j_q', 'j_w', 'j_z', 'j_t', 'j_index'], 'KerasModel': 'conv2d_model', 'KerasModelRetrain': 'conv2d_model_constraint', 'KerasLoss': 'categorical_crossentropy', 'L1Reg': 0.0001, 'NormalizeInputs': 0, 'InputType': 'Conv2D', 'BinsX': 40, 'MinX': -0.8, 'MaxX': 0.8, 'BinsY': 40, 'MinY': -1.0, 'MaxY': 1.0, 'MaxParticles': 100}\n"
     ]
    }
   ],
   "source": [
    "from collections import namedtuple\n",
    "Option = namedtuple(\"MyStruct\", \"inputModel inputFile tree config jsonModel\")\n",
    "\n",
    "options = Option(\n",
    "    #inputModel = '../weights/KERAS_conv2d_model_weights.h5',\n",
    "    inputModel = '../conv2d_weights/KERAS_check_best_model.h5',\n",
    "    inputFile = '../data/processed-pythia82-lhc13-all-pt1-50k-r1_h022_e0175_t220_nonu_withPars_truth.z',\n",
    "    tree = 't_allpar_new',\n",
    "    config = 'train_config_conv2d.yml',\n",
    "    #jsonModel = '../weights/KERAS_conv2d_model.json'\n",
    "    jsonModel = '../conv2d_weights/model.json'\n",
    ")\n",
    "\n",
    "print(\"Loading configuration from\", options.config)\n",
    "config = open(options.config, 'r')\n",
    "yamlConfig =  yaml.load(config, Loader=yaml.FullLoader)\n",
    "print(yamlConfig)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Keys: <KeysViewHDF5 ['t_allpar_new']>\n",
      "load 24816\n",
      "load 0\n",
      "load 1000\n",
      "load 2000\n",
      "load 3000\n",
      "load 4000\n",
      "load 5000\n",
      "load 6000\n",
      "load 7000\n",
      "load 8000\n",
      "load 9000\n",
      "load 10000\n",
      "load 11000\n",
      "load 12000\n",
      "load 13000\n",
      "load 14000\n",
      "load 15000\n",
      "load 16000\n",
      "load 17000\n",
      "load 18000\n",
      "load 19000\n",
      "load 20000\n",
      "load 21000\n",
      "load 22000\n",
      "load 23000\n",
      "load 24000\n",
      "(24816, 40, 40, 1)\n",
      "(19852, 40, 40, 1)\n",
      "(4964, 40, 40, 1)\n"
     ]
    }
   ],
   "source": [
    "X_train_val, X_test, y_train_val, y_test, labels  = get_features(options, yamlConfig)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_sum = np.zeros((5, 40, 40))\n",
    "for i in range(X_train_val.shape[0]):\n",
    "    X = X_train_val[i].squeeze() # (40, 40)\n",
    "    train_sum[np.argmax(y_train_val[i])] += X\n",
    "#train_sum /= X_train_val.shape[0]\n",
    "train_sum /= np.ndarray.max(train_sum)\n",
    "for i in range(train_sum.shape[0]):\n",
    "    XB = np.linspace(0,40,40)\n",
    "    YB = np.linspace(0,40,40)\n",
    "    X,Y = np.meshgrid(XB,YB)\n",
    "    Z = train_sum[i]\n",
    "    plt.imshow(Z,interpolation='none')\n",
    "    plt.show()\n",
    "    #hist, xedges, yedges = np.histogram2d(train_sum, y, weights=w, bins=(xbins,ybins))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 40, 40, 1)         0         \n",
      "_________________________________________________________________\n",
      "conv1_relu (Conv2D)          (None, 40, 40, 8)         976       \n",
      "_________________________________________________________________\n",
      "conv2_relu (Conv2D)          (None, 20, 20, 4)         292       \n",
      "_________________________________________________________________\n",
      "conv3_relu (Conv2D)          (None, 10, 10, 2)         74        \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                6432      \n",
      "_________________________________________________________________\n",
      "output_softmax (Dense)       (None, 5)                 165       \n",
      "=================================================================\n",
      "Total params: 7,939\n",
      "Trainable params: 7,939\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "json_file = open(options.jsonModel, 'r')\n",
    "model = json_file.read()\n",
    "json_file.close()\n",
    "model = model_from_json(model, custom_objects={'ZeroSomeWeights':ZeroSomeWeights,\n",
    "                                                           'BinaryDense': BinaryDense,\n",
    "                                                           'TernaryDense': TernaryDense,\n",
    "                                                           'QuantizedDense': QuantizedDense,\n",
    "                                                           'binary_tanh': binary_tanh,\n",
    "                                                           'ternary_tanh': ternary_tanh,\n",
    "                                                           'quantized_relu': quantized_relu,\n",
    "                                                           'Clip': Clip})\n",
    "# load weights into new model\n",
    "model.load_weights(options.inputModel)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.environ[\"PATH\"] += os.pathsep + 'C:/Program Files (x86)/Graphviz2.38/bin/'\n",
    "from keras.utils.vis_utils import plot_model\n",
    "plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)\n",
    "from IPython.display import Image\n",
    "#Image(filename='model_plot.png') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using the weights provided"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot ROC curve\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Error when checking : expected conv2d_1_input to have shape (8, 8, 1) but got array with shape (40, 40, 1)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-14-bae2123f3d9e>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Plot ROC curve\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0my_predict\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmakeRoc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_test\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlabels\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my_test\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshow\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Plot confusion matrix\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\JupyterNotebooks\\keras-training\\train\\eval.py\u001b[0m in \u001b[0;36mmakeRoc\u001b[1;34m(features_val, labels, labels_val, model)\u001b[0m\n\u001b[0;32m     71\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[1;34m'j_index'\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mlabels\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mlabels\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mremove\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'j_index'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     72\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 73\u001b[1;33m     \u001b[0mpredict_test\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfeatures_val\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     74\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     75\u001b[0m     \u001b[0mdf\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mDataFrame\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\keras\\models.py\u001b[0m in \u001b[0;36mpredict\u001b[1;34m(self, x, batch_size, verbose, steps)\u001b[0m\n\u001b[0;32m   1023\u001b[0m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbuild\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1024\u001b[0m         return self.model.predict(x, batch_size=batch_size, verbose=verbose,\n\u001b[1;32m-> 1025\u001b[1;33m                                   steps=steps)\n\u001b[0m\u001b[0;32m   1026\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1027\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mpredict_on_batch\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mx\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\keras\\engine\\training.py\u001b[0m in \u001b[0;36mpredict\u001b[1;34m(self, x, batch_size, verbose, steps)\u001b[0m\n\u001b[0;32m   1815\u001b[0m         x = _standardize_input_data(x, self._feed_input_names,\n\u001b[0;32m   1816\u001b[0m                                     \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_feed_input_shapes\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1817\u001b[1;33m                                     check_batch_axis=False)\n\u001b[0m\u001b[0;32m   1818\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstateful\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1819\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mx\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m>\u001b[0m \u001b[0mbatch_size\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mx\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m%\u001b[0m \u001b[0mbatch_size\u001b[0m \u001b[1;33m!=\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\keras\\engine\\training.py\u001b[0m in \u001b[0;36m_standardize_input_data\u001b[1;34m(data, names, shapes, check_batch_axis, exception_prefix)\u001b[0m\n\u001b[0;32m    121\u001b[0m                             \u001b[1;34m': expected '\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mnames\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m' to have shape '\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    122\u001b[0m                             \u001b[0mstr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m' but got array with shape '\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 123\u001b[1;33m                             str(data_shape))\n\u001b[0m\u001b[0;32m    124\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0mdata\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    125\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mValueError\u001b[0m: Error when checking : expected conv2d_1_input to have shape (8, 8, 1) but got array with shape (40, 40, 1)"
     ]
    }
   ],
   "source": [
    "print(\"Plot ROC curve\")\n",
    "y_predict = makeRoc(X_test, labels, y_test, model)\n",
    "plt.show()\n",
    "\n",
    "print(\"Plot confusion matrix\")\n",
    "y_test_proba = y_test.argmax(axis=1)\n",
    "y_predict_proba = y_predict.argmax(axis=1)\n",
    "\n",
    "# Compute non-normalized confusion matrix\n",
    "cnf_matrix = confusion_matrix(y_test_proba, y_predict_proba)\n",
    "np.set_printoptions(precision=2)\n",
    "# Plot non-normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=[l.replace('j_','') for l in labels],\n",
    "                          title='Confusion matrix')\n",
    "plt.show()\n",
    "\n",
    "print(\"Plot normalized confusion matrix\")\n",
    "# Plot normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=[l.replace('j_','') for l in labels], normalize=True,\n",
    "                          title='Normalized confusion matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using the weight I trained"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot ROC curve\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEXCAYAAACzhgONAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3deXwUVbbA8d8hBsMS2VRkcQgYtgRI2EHZIYDIMoyjgij4FEEUYZwZR1ERt3FFnzKOIIoCgoCiCLiBM4IBFYEIAqIgOOhEIiBrWAJZzvujO/2ydDqVpZNO9/l+Pv2h61Z11bkB+qTqVp0rqooxxhhTkErlHYAxxpjAZonCGGOMT5YojDHG+GSJwhhjjE+WKIwxxvhkicIYY4xPQZ8oRGSfiKiIzC3JNmVNRB5yx2T3LxtjylXQJ4qyICLPZH+pi8jm8o7HGGNKkyWKEhKRPsBfyjsOY4zxl1BKFCIi94nIfhE5KiILRCTSx8Z3ichOETklIidE5FsReT3PNrWB+cCPwNcF7Cf7TONpEZkjIidF5EcRGS4ijUXkX+5jbBWRLqXaY2OMKQXnlXcAZegaIAP4DagHjAJ+Au7Pu6GIDAGecy9+BygQBTQH/ifHprOBukA34KlCjj/JfexzQGPgTSAlx/o4YLGIRKtqRhH6ZYwxfhVKZxQZQEsgGsgeR+hbwLbN3H/+W1VjVDUWqAH0yd5ARG4BrgYeUtWvHBx/D3AZcK17OQL4wd02yd3WyB2fMcYEjFBKFJ+q6i+qmgXscrfVLWDbVbh+8+8rIr+JyBfADCAdQEQuBZ4HEoEnHB5/taqeBfblaPtAXVUZf8zRVlBMxhhTLkLp0tOxHO+zL+2Itw1VdYeIxALXA21xXRaaAIxzjyNUd786AydEBKCK++NtReQk0FVVt+fY7Yk8x87ZlvMWWK8xGWNMeQmlROGYiDQFslT1EfdyBHAYqAr0BJLcm57vfuVUCagGhJVNtMYY41+hdOmpKHoCe9x3SH0N7MWVJAC2qepaVZWcL+Az9/okd9vW8gjcGGNKmyUK77YA7wJncQ2AR7rbblHVT8ozMGOMKWtiM9wZY4zxxc4ojDHG+BTwg9kiUg14CdftqmtVdWE5h2SMMSGlXM4oROQ1ETkoIjvytA8UkV0iskdE7nU3/wFYqqq3AkPLPFhjjAlx5XXpaS4wMGeDiIQB/wSuBGKAkSISAzQE/uveLLMMYzTGGEM5XXpS1UQRicrT3AnYo6o/AojIYmAYkIwrWWzFR2ITkXHAOIAqVaq0v/TSS4sVW1ZWFpUqhdbQjfU5NFifg9vZw2mgUKl2BOHF7PLu3bt/U9WL8rYH0hhFA/7/zAFcCaIzrtIZL4rIVcDKgj6sqrNxFemjQ4cOunlz8aaFWLt2Lb169SrWZysq63NosD4HtzljXgaF3s+OoslF1Yu1DxH5yVt7ICUKb6UrVFVPkbtiqzHGmDIUSOdkyUDO60UNgf3lFIsxxhi3QEoUm4Cm7sl8KgMjgBVF2YGIDBGR2cePH/dLgMYYE4rK5dKTiCwCegEXikgyME1V54jIRFwlvsOA11T126LsV1VXAis7dOhwa9516enpJCcnk5aW5nMfNWrU4LvvvivKYSs867N/RURE0LBhQ8LDw8vkeMaUtvK662lkAe0fAh/645jJyclERkYSFRWFuyy4V6mpqURGFjhDalCyPvuPqnL48GGSk5Np3Lix349njD8E0qWnEvN16SktLY06der4TBLGlDYRoU6dOoWeyRoTyIIqUajqSlUdV6NGDa/rLUmY8mD/7kxFF1SJwhhjTOmzRBFg3nvvPXbu3FneYfg0efJkGjRoQFZWlqftoYceYvr06bm2i4qK4rfffgPg119/ZcSIEVx22WXExMQwaNAgdu/eXaI4EhMTadeuHeeddx5Lly7NtW7evHk0bdqUpk2bMm/ePE97UlISrVu3Ji4ujkmTJuGtzL6qMmnSJKKjo2nTpg1ff/01ALt27aJ9+/bExcXx5ZdfApCRkUG/fv04ffp0ifpiTCCzRBFgyjtRZGRk+FyflZXFsmXLuPTSS0lMTHS0T1Vl+PDh9OrVi71797Jz504ef/xxDhw4UKJYf/e73zF37lyuv/76XO1Hjhzh4Ycf5quvvmLjxo08/PDDHD16FIAJEyYwe/Zstm7dyg8//MDHH3+cb78fffQRP/zwAz/88AOzZ89mwoQJALz88ss8+eSTLF261JMUZ86cyY033kjVqlXz7ceYYBFUiSLQn6N49NFHadGiBQkJCYwcOTLfb+BffPEFK1as4O677yY+Pp69e/fyyiuv0LFjR+Li4rj66qs9v7nu3buXLl260LFjRx588EGqV3c9sp+VlcXtt99ObGwsgwcPZtCgQZ7ftpOSkujZsyft27dnwIABpKSkADBo0CDuu+8+evbsyQsvvOCzD2vWrKFVq1ZMmDCBRYsWOer3mjVrCA8P57bbbvO0xcfH0717d2c/uAJERUXRpk2bfLV8Vq1aRUJCArVr16ZWrVokJCTw8ccfk5KSwokTJ+jatSsiwujRo3nvvffy7Xf58uWMHj0aEaFLly4cO3aMlJQUwsPDOXPmDKdPnyY8PJxjx46xcuVKRo8eXaJ+GBPoAqmER4n5eo4ip4dXfsvO/Se8rsvMzCQsLKzIx46pfwHThsQWuH7z5s288847bNmyhYyMDNq1a0f79u1zbXP55ZczdOhQBg8ezB//+EcAatasya23urrzwAMPMGfOHO68804mT57M5MmTGTlyJLNmzfLs491332Xfvn1s376dgwcP0rJlS26++WbS09O58847Wb58ORdddBFLlizh/vvv57XXXgPg2LFjfPbZZxRm0aJFjBw5kmHDhnHfffeRnp5e6PMBO3bsyNfXgnTv3p3U1NR87dOnT6dfv36O9vHLL7+Qsyhkw4YN+eWXX/jll19o2LBhvnann7/jjjsYPXo0Z8+e5eWXX+aRRx7h/vvvt8FqE/SCKlEEsvXr1zNs2DCqVKkCwJAhQxx9bseOHTzwwAMcO3aMkydPMmDAAAC+/PJLz2/D119/PX/96189x7nmmmuoVKkSl1xyCb179wZc19d37NhBQkIC4EqI9erV8xznuuuuKzSWc+fO8eGHH/K///u/REZG0rlzZ1avXs1VV11V4JdlUb9E161bV6TtvfE27iAiBbY7/fzvfvc71q5dC8CePXvYv38/LVq04MYbb+TcuXM8+uijNGvWrMTxGxNoQjJR+PrN318PYhV3bvKbbrqJ9957j7i4OObOnev5oirqcVSV2NhYzyBsXtWqVSs0lo8//pjjx4/TunVrAE6fPk3VqlW56qqrqFOnjudSVrbU1FRq1qxJbGxsvsHmgpTGGUXDhg1z/ZySk5Pp1asXDRs2JDk5OVd7/fr1vX7+v//9r8/t7r//fh577DFmzJjBqFGjiIqK4uGHH2bhQpuA0QQfG6MoI926dWPlypWkpaVx8uRJPvjgA6/bRUZG5vqiTE1NpV69eqSnp+f6EurSpQvvvPMOAIsXL851nHfeeYesrCwOHDjg+cJs3rw5hw4d8iSK9PR0vv3We4WUZcuWMWXKlHztixYt4tVXX2Xfvn3s27eP//znP6xevZrTp0/To0cPVqxY4Yn93XffJS4ujrCwMPr06cPZs2d55ZVXPPvatGmT10td69atY+vWrfleTpMEwIABA1i9ejVHjx7l6NGjrF69mgEDBlCvXj0iIyPZsGEDqsr8+fMZNmxYvs8PHTqU+fPno6ps2LCBGjVq5Dr7+uyzz2jQoAFNmzbl9OnTVKpUibCwMLvzyQQvVQ26V/v27TWvnTt35mvz5sSJE462K45p06Zps2bNNCEhQa+//nqdPXt2vm3Wr1+vLVu21Pj4eN2zZ4++9NJLGhUVpT179tSJEyfqmDFjVFV19+7d2qlTJ+3YsaM+9NBDWr9+fVVVzczM1PHjx2vLli112LBhOnDgQF29erWqqm7ZskW7d++ubdq00ZiYGM/xu3Xrpps2bfLE8Mwzz+jjjz+eK65Tp05prVq19Pjx47nahw8frosXL1ZV1VmzZmmbNm00Li5OExISdO/evZ7tfvnlF73mmmu0SZMmGhMTo4MGDdLdu3eX6Oe5ceNGbdCggVatWlVr166tMTExnnVz5szRyy67TC+77DJ97bXXPO2bNm3S2NhYjYqK0jvuuEOzsrJUVXXmzJk6c+ZMVVXNysrS22+/XZs0aaKtWrXK9bPJysrSfv366ZEjR1TV9e+qbdu22rp1a12/fn2BsTr99+dPa9asKe8Qylwo9fnV0bP01Rtn6d6DqcXeB7BZvXynlvuXuj9egZooUlNdf4GnTp3S9u3ba1JSUrH3derUKc+X3KJFi3To0KH5jvPbb79pkyZNNCUlxee+8vZ51KhRevDgwWLHVhH48+/ZG0sU5SOU+uzPRBGSYxTlZdy4cezcuZO0tDTGjBlDu3btir2vpKQkJk6ciKpSs2ZNz91LAIMHD+bYsWOcO3eOqVOncskllxRp3wsWLCh2XMaY4GOJogy9+eabpbav7t27880333hdV9iAtzHGFEVQDWYbY4wpfUGVKAL5ridjjKmogipRaCFlxo0xxhRdUCUKY4wxpc8SRYAp7+qxTliZcSszbkKLJYoAU96JwsqMW5lxY/KyRFGG/v73v9O8eXP69etnZcatzLgxFUZoPkfx0b3w63avq6pkZkBYMX4sl7SGK58scHVSUhKLFy+2MuOFsDLjxgSeoEoUIjIEGBIdHV3eoeSzbt06hg8f7rlEMXToUEefszLjRedt3MHKjBtTfEGVKNThxEW+fvM/46cy41D0L02wMuPZrMy4MeXHxijKSI8ePVi2bBlnzpwhNTWVlStXet3OyoxbmXFjAk1QnVEEsnbt2nHdddcRHx9Po0aNChzIHTFiBLfeeiszZsxg6dKlPProo3Tu3JlGjRrRunVrzxfx888/zw033MCzzz7LVVddRfZDhldffTX//ve/adWqFc2aNaNz587UqFGDypUrs3TpUiZNmsTx48fJyMjgT3/6E7Gx+Sdx2rt3LxdccEGuttOnT7Nq1SpefvllT1u1atU882xcd911TJw4kW7duiEiXHzxxbz66quA60xq2bJl/OlPf+LJJ58kIiKCqKgonn/++RL9TDdt2sTw4cM5evQoK1euZNq0aXz77bfUrl2bqVOn0rFjRwAefPBBateuDbjuUrrppps4deoUV111FVdeeSWAZ5zntttuY9CgQXz44YdER0dTtWpVXn/9dc8xVZXHHnuMt956C3AVehw1ahQZGRnMnDmzRP0xJmB5Kylb0V+BWmY8p2nTpukzzzxT7M9bmfGSsTLjoSGU+mxlxk0+VmbcGFNWLFGUk4ceeqhEn7cy48aYsmKD2cYYY3yyRGGMMcanoEoUNh+FMcaUvqBKFGrzURhjTKkLqkQRTJ5//vmAf4ArLi6OkSNH5mrr1asXmzdv9izv27ePVq1aeZY3btxIjx49aN68OS1atGDs2LEl7ueLL75IdHQ0IuIpaw4FlwsHWLhwodcy5DkdOXKEhIQEmjZtSkJCgqcC7TvvvENsbCzdu3fn8OHDgOvZkxEjRpSoH8YEKksUAaq8E0VmZqbP9d999x1ZWVkkJiZy6tQpR/s8cOAA11xzDU899RS7du3iu+++Y+DAgV5LdhTFFVdcwb/+9S8aNWqUq72gcuFHjhzhqaee8lqGPKcnn3ySvn378sMPP9C3b1+efNJV+uXZZ59lw4YNjB49mjfffBNwFWx89NFHS9QPYwKVJYoy8vTTTzNjxgwA7rrrLvr06QPAv//9b2644YZc286YMYP9+/fTu3dvT1G/CRMm0KFDB2JjY5k2bZpn2w8//JAWLVrQrVs3Jk2axODBgwE4dOgQCQkJtGvXjvHjx9OoUSPPb9sLFiygU6dOxMfHM378eE9SqF69Og8++CCdO3cusCZUtjfffJMbb7yR/v37s2LFCkc/g3/+85+MGTOGrl27Aq4ntv/4xz9St25dR58vSNu2bYmKisrXXlC58FWrVtG7d+98Zci9fX7MmDEAjBkzxlOEsVKlSpw9e9ZTbnzdunXUq1ePpk2blqgfxgSqkHyO4qmNT/H9ke+9rsvMzCQsLKzI+2xRuwX3dLqnwPU9evTg2WefZdKkSWzevJmzZ8+Snp7O+vXr85XzmDRpEs899xxr1qzhwgsvBFxzWdSuXZvMzEz69u3Ltm3baNasGePHjycxMZHGjRvnugz08MMP06dPH6ZMmcLHH3/M7NmzAdeZwJIlS/j8888JDw/n9ttvZ8mSJYwfP55Tp07RqlUrHnnkkUL7u2TJEj755BN27drFiy++mO8SlDc7duzwfPH6smvXrgKr2a5du5aaNWsWug/wXW68QYMG+drzOnDggKfGU7169Th48CAA06ZNY8CAAdSvX58FCxZw7bXX5qq3ZUywCclEUR7at29PUlISqampnH/++bRr147Nmzezbt06z5mGL2+99RazZ88mIyODlJQUdu7cSVZWFk2aNKFx48YAjBw50pMQ1q9fz7JlywAYOHAgtWrVAlxnMElJSZ46SGfOnPHUiQoLC+Pqq68uNJZNmzZx0UUX0ahRIxo2bMjNN9/M0aNHqVWrltcKuUWtmtu8eXO2bt1apM94oyUsN16QhIQET7n2efPmMWjQIHbt2sX06dOpVasWL7zwgs14Z4JKSCYKX7/5p/qpzHh4eDhRUVG8/vrrXH755bRp04Y1a9awd+9eWrZs6fOz//nPf5g+fTqbNm2iVq1a3HTTTaSlpRVYUhx8lxsfM2YMTzzxhKcte4wgIiLC0dnUokWL+P777z2Xe06cOME777zD2LFjqVOnTq7r/UeOHPGcFcXGxpKUlOS1YmtOpXVGUVC58IYNG/LJJ5/kau/Vq1e+z9etW5eUlBTq1atHSkoKF198ca71p0+fZt68eaxatYr+/fuzfPly3nzzTRYuXOiZbMqYYGBjFGWoR48eTJ8+nR49etC9e3dmzZpFfHy8199mc5YbP3HiBNWqVaNGjRocOHCAjz76CIAWLVrw448/sm/fPsB1OShbt27dPBVOs0tuA/Tt25elS5d6LqMcOXKEn3/+2Wu8U6ZM8ZyVZMvKyuLtt99m27ZtnnLjy5cv90yL2qtXLxYsWOBJVPPmzfOMs0ycOJF58+bx1Vdfefa3YMECfv3111zHyD6j8PZymiSg4HLhAwYM4NNPP81Xhtzb57PviJo3b16+BPf0008zefJkzxSpIkKlSpUC/m41Y4rKEkUZ6t69OykpKXTt2pW6desSERFRYLnxcePGceWVV9K7d2/i4uJo27YtsbGx3HzzzVxxxRUAVKlShZdeeomBAwfSrVs36tat67mMNG3aNFavXk27du346KOPPHMxxMTE8Nhjj9G/f3/atGlDQkJCvi/qbNu3b89XUDAxMZEGDRrkusbfo0cPdu7cSUpKCuPGjSMyMpK4uDji4uI4efKkZ/a9unXrsnjxYv7617/SvHlzWrZsybp16/KVNC+qGTNmeCYlatOmDWPHjgVcc4E3adKE6Ohobr31Vl566SUAateuzd/+9jc6duzomXM8uwz52LFjPbf33nvvvXzyySc0bdqUTz75hHvvvddzzP3797N582ZP8vjLX/5Cly5dmDdvHtdff32J+mNMwPFWUraivypCmfHSkl1SPCsrSydMmKDPPfecqqqmpaVpenq6qqp+8cUXGhcXV+A+Cupz//79SznawGFlxkNDKPXZyoybAr3yyivMmzePc+fO0bZtW8aPHw/Azz//zLXXXktWVhaVK1fONbucU6tWrSrtcI0xFZAligrurrvu4q677srX3rRpU7Zs2VIOERljgk1QjVFYUUBjjCl9QZUo1IoCGmNMqQuqRGGMMab0FZooRKRVYdsYY4wJXk7OKGaJyEYRuV1EnD/tZIqsvCvGOlGRS4vPmzfPU1p84cKFXvdrpcWNya/QRKGq3YBRwKXAZhF5U0QS/B5ZCCrvRBHspcUffvhhT2nxp556ykqLG+OQozEKVf0BeAC4B+gJzBCR70XkD/4MLphkl+uIj4+ncePGnrIW2cqztPhjjz0WEqXFExISPKXFe/fubaXFjXGo0OcoRKQN8D/AVcAnwBBV/VpE6gNfAu/6N8TS9+vjj3P2O+9lxjMyMzlSjDLj57dswSX33Vfg+ttuu43bbruN9PR0+vTpw5///Odc68u6tPjChQsZPXo0p06dIiYmhqeeeqrQPlb00uI52+vXr2+lxY1xyMkDdy8CrwD3qeqZ7EZV3S8iD/gtsiA1efJk+vTpw5AhQwrd1p+lxbMroYaFhRVazRWstLiVFjehzEmiGAScUdVMABGpBESo6mlVfcOv0fmJr9/8/VVmHGDu3Ln89NNPvPjii4Vu6+/S4tlCqbT42rVrPe379+8nNjY23+ettLgx+TkZo/gXUCXHclV3mymCpKQkpk+fzoIFC6hUyfuPvSxLi//0009eYwjm0uLZP5OjR4/y6aefWmlxYxxykigiVPVk9oL7vZ1jF9GLL77IkSNH6N27N/Hx8Z5S2DmVZWnxlJQUr3EGc2nxqVOnekqL33PPPVZa3BinvJWUzfkCPgfa5VhuD3xZ2OfK8xXsZcZLo7R4Tjn7HMylxXOyMuOhIZT6XN5lxv8EvC0i+93L9QDvF5NNmbDS4saYslRoolDVTSLSAmgOCPC9qqb7PTJTICstbowpS07no+gIRLm3b+u+3XC+36IyxhgTMJw8cPcGcBmwFciu8aCAJQpjjAkBTs4oOgAx7oEOY4wxIcbJ7bE7gEsK3cpPRKSJiMwRkaXlFYMxxoQyJ4niQmCniKwSkRXZLyc7F5HXROSgiOzI0z5QRHaJyB4RubegzwOo6o+qeouT4wWyY8eOee7pL8q6QHHo0CHCw8N5+eWXc7VXr1491/LcuXOZOHGiZ3n+/Pm0atWK2NhYYmJimD59eoljufnmm7n44otzlTKHgkuEq4/S4zklJSXRunVroqOjmTRpkucBwnvuuYc2bdowevRoz7ZvvPEGL7zwQon7YkxF4CRRPAT8HngceDbHy4m5wMCcDSISBvwTuBKIAUaKSIyItBaR9/O8Ls6/y4opkBNFYeXFAd5++226dOnieRrbiY8++ojnn3+e1atX8+233/L1119TGtPU3nTTTV4rvxZUIryg0uN5TZgwgdmzZ3u2/fjjjzl+/DhffPEF27ZtIzMzk+3bt3PmzBnmzp3L7bffXuK+GFMROJmP4jNgHxDufr8J8P4rWf7PJgJH8jR3Ava4zxTOAYuBYaq6XVUH53kdLEpnAtm9997L3r17iY+P5+677/a57uTJk/Tt25d27drRunVrli9f7tn20UcfpUWLFiQkJDBy5EjPb+ibNm2iTZs2dO3albvvvtvz23ZmZiZ33303HTt2pE2bNp4zgrVr19K7d2+uv/56unTpUmj8ixYt4tlnnyU5Odlr1VVvnnjiCaZPn079+vUBV12p0qiH1KNHD89T1TkVVCLcW+nxvCVEUlJSOHHiBF27dkVEGD16NO+99x6VKlXi3LlzqCpnzpwhPDycZ555hkmTJhEeHl7ivhhTETi56+lWYBxQG9fdTw2AWUDfYh6zAfDfHMvJQGcfx68D/B3XbblTVDV/ZTvXduPccVK3bt1cBeAAatSo4amjtGn5zxzZ7702j6oWufIpQO36Vek47HcFrn/ggQfYtm0b69atA8g1cU/edenp6cyfP58LLriAw4cP06dPH3r37s2WLVt4++23SUxMJCMjg+7du9OqVStSU1MZM2YMM2bMoHPnzkybNo2srCxSU1N5/fXXiYiI4NNPP+Xs2bP079+fyy+/nNOnT7Nx40Y2bNjApZde6nMioeTkZPbv30/Lli35/e9/z/z583NdXsr52bS0NM6dO0dqairbt2+nWbNmhU5StGTJEmbMmJGvvUmTJrzxhve6kydPnvT0MduBAweoXr06qampVK9enYMHD5KamspPP/1EnTp1PNvWq1eP5OTkXKVKdu/eTb169Tzb1K5d21MPa/DgwcTFxdGzZ0/CwsL48ssvueuuu4o0+VJaWlq+f5Nl7eTJk+UeQ1kLxT5v3LiRn6s5mmrIMSd3Pd2B6yzgK3BNYlTCS0LevoULvKNKVQ8DtxW2U1WdDcwG6NChg/bq1SvX+u+++85TFTa8cniBFVMzMzMdVVPNK7xyuM+qs9WrV6dSpUpet8m7Lj09nalTp5KYmEilSpVISUnh9OnTbNmyheHDh3sqmg4bNozzzz+fzMxMTp06Rb9+/QDXpZnVq1cTGRlJYmIi27ZtY+XKlQAcP36clJQUqlatSqdOnWjdunWhFXPff/99RowYQWRkJKNHj+aWW25hypQpnvU5PxsREcH5559PZGQkIkJkZGSh1XjHjh3rtfaVLwX9PL0th4WFUbVqVc+6sLAwwsLCcm1btWrVXG1Vq1YlPNz1dzp16lSmTp3qifXxxx9nyZIlrF69mjZt2vDAA4VX24+IiKBt27ZF6mNpW7t2LXn/XwS7UOrz3td3AdCpUyeaXFS9kK2LxkmiOKuq57J/yxaR8/Dxxe5AMq5pVbM1BPYXsK1fdL+2WYHr/Flm3KmFCxdy6NAhkpKSCA8PJyoqymeJcV93Lqsq//jHP/JVSl27di3VqlVzFM+iRYs4cOCAZ57p/fv388MPP9C0aVOqVKnCuXPnqFy5MuC9xHifPn0K7e8zzzyTrz06OpqlS53f7FZQiXBvpcezJyfKll1gMOc22ZfMsmU/9d6sWTMmT55MYmIiI0aM8PwsjAlWTs5PPhOR+4Aq7rmy3wZWluCYm4CmItJYRCoDIwBn82oWQkSGiMjs48ePl8buSlXOEuKFrTt+/DgXX3wx4eHhrFmzxnMJpFu3bqxcuZK0tDROnjzJBx98AECtWrWIjIxkw4YNALlmXhswYAAzZ84kPd1VdWX37t0Fznfdt2/ffOMPu3bt4tSpU/zyyy+eEuNTpkzxHKNnz54sWLAAcE2I9NZbb3lKjE+ZMoW//e1vnvGAs2fPer3ENGrUKK/lxYuSJKDgEuHeSjSZrWcAABSMSURBVI/nrZCbXWl3w4YNqCrz58/PV2J86tSpPPLII6Snp3tuALAS4yYUOEkU9wKHgO3AeOBDXPNnF0pEFuGaLrW5iCSLyC2qmgFMBFYB3wFvqeq3xQk+L1VdqarjSuPOmtJWp04drrjiClq1apVvMDvvulGjRrF582Y6dOjAwoULadGiBQAdO3Zk6NChxMXF8Yc//IEOHTp47iKaM2cO48aNo2vXrqiqp33s2LHExMTQrl07WrVqxfjx48nIyMgXX1ZWFnv27Mk3SLxo0SKGDx+eq+3qq6/23P30wgsv8O677xIfH0+XLl245ppr6NGjB+Aq+X3HHXfQr18/YmNjad++vddjF9XIkSPp2rUru3btomHDhsyZMwcouER4QaXHAeLj4z3vZ86cydixY4mOjuayyy7jyiuv9Kx777336NixI/Xr16dmzZp07dqV1q1bIyLExcWVuE/GBDRvJWUr+iuYy4xnlxg/deqUtm/fXpOSknK1q6o+8cQTOmnSJMf7PHHihG7fvl3vuuuu0g02gFmZ8dAQSn0ulzLjIvKWql4rItvxMiahqm38mcCMd+PGjWPnzp2kpaUxZswY2rVrB8AHH3zAE088QUZGBo0aNWLu3LlF2m+rVq147rnn/BCxMaai8zWYPdn95+CyCKQ0iMgQYEh0dHR5h+I3b775ptf26667rsA5p40xpiQKHKNQ1ZQc2xxQ1Z9U9SfgIN5vcS13WsgYhVpdQ1MO7N+dqeicDGa/DWTlWM50t1UoERERHD582P7TmjKlqhw+fJiIiIjyDsWYYnPyHMV56iq1AYC6nqmo7MeY/CL7PvlDhw753C4tLS3k/lNbn/0rIiKChg0blsmxjPEHJ4nikIgMVdUVACIyDPjNv2EVj68xivDwcBo3blzoPtauXVvuT9CWNeuzMcYXJ5eebgPuE5GfReS/wD24nqcIOIWNURhjjCm6Qs8oVHUv0EVEqgOiqs4roRljjKnwfD1HcYOqLhCRP+dpB0BV7aZ7Y4wJAb7OKKq6/yzfCnnGGGPKla9EcZn7z52qWiFuhw2FB+6MMaas+RrMHiQi4cAUH9sEFBvMNsaY0ufrjOJjXLfBVhOREznaBVBVvcCvkRljjAkIvs4oHlDVGsAHqnpBjlekJQljjAkdvhLFl+4/T/jYxhhjTJDzdempsoiMAS4XkT/kXamq7/ovLGOMMYHCV6K4DRgF1ASG5FmnQMAlCrvryRhjSl+BiUJV1wPrRWSzqs4pw5iKTVVXAis7dOhwa3nHYowxwaLAMQoR+RuAqs4RkWvyrHvc34EZY4wJDL4Gs0fkeJ/3WYqBfojFGGNMAPKVKKSA996WjTHGBClfiUILeO9t2RhjTJDydddTnPuJbAGq5Hg6W4DQmg7NGGNCmK+7nsLKMpDSYLfHGmNM6XMyw12FYUUBjTGm9AVVojDGGFP6LFEYY4zxyRKFMcYYn3zNmZ2Kj9tgrdS4McaEBl93PUUCiMgjwK/AG7hujR2FzaNtjDEhw8mlpwGq+pKqpqrqCVWdCVzt78CMMcYEBieJIlNERolImIhUEpFRQKa/AzPGGBMYnCSK64FrgQPu1zXutoAjIkNEZPbx48fLOxRjjAkahSYKVd2nqsNU9UJVvUhVf6+q+8ogtiKzB+6MMab0+ar1BICIXATcCkTl3F5Vb/ZfWMYYYwJFoYkCWA6sA/6FjU0YY0zIcZIoqqrqPX6PxBhjTEByMpj9vogM8nskxhhjApKTRDEZV7I4IyInRCQ1x9wUxhhjglyhl56yn9A2xhgTmpzc9dTDW7uqJpZ+OMYYYwKNk8Hsu3O8jwA6AUlAH79EZIwxJqA4ufQ0JOeyiFwKPO23iIwxxgSU4sxHkQy0Ku1AjDHGBCYnYxT/4P/npagExAPf+DMoY4wxgcPJGMXmHO8zgEWq+rmf4ikRERkCDImOji7vUIwxJmg4KQo4D1iEawD7G2Cjv4MqLisKaIwxpc/JpadewDxgH64Z7i4VkTF2e6wxxoQGJ5eengX6q+ouABFphusMo70/AzPGGBMYnNz1FJ6dJABUdTcQ7r+QjDHGBBJHg9kiMgd4w708Ctd4hTHGmBDgJFFMAO4AJuEao0gEXvJnUMYYYwKHz0QhImHAHFW9AXiubEIyxhgTSHyOUahqJnCRiFQuo3iMMcYEGCeXnvYBn4vICuBUdqOq2hmGMcaEACeJYr/7VQmwuSmMMSbEOKke+3BZBGKMMSYwOXkyeyX/XxQw23FcNaBeVtU0fwRmjDEmMDh54O5H4CTwivt1AjgANHMvG2OMCWJOxijaqmrO6VBXikiiqvYQkW/9FZgxxpjA4OSM4iIR+V32gvv9Re7Fc36JyhhjTMBwckbxF2C9iOzF9WR2Y+B2EamGq6qsMcaYIObkrqcPRaQp0AJXovje1axngef9HJ8xxphyVuilJxF5TVXPquo3qroVCAM+9H9onuP/XkReEZHlItK/rI5rjDHGxckYxS8iMhNARGoBnwALnOxcRF4TkYMisiNP+0AR2SUie0TkXl/7UNX3VPVW4CbgOifHNcYYU3qcTIU6FTghIrOA1cCzqvq6w/3PBQbmbHAXGvwncCUQA4wUkRgRaS0i7+d5XZzjow+4P2eMMaYMiWreZ+ncK0T+kHMRmIprvuyPAVT1XUcHEIkC3lfVVu7lrsBDqjrAvTzFvb8nCvi8AE8Cn6jqv3wcZxwwDqBu3brtFy9e7CS8fE6ePEn16tWL9dmKyvocGqzPwW3v67tAodp1TbmkmpOLRfn17t07SVU75G33NZg9JM/yFlwz2w3B9aS2o0ThRQPgvzmWk4HOPra/E+gH1BCRaFWd5W0jVZ0NzAbo0KGD9urVq1jBrV27luJ+tqKyPocG63Nw2/u6ayLSTp060eSi0k2OBSYKVf2fUj3S/xNvh/MRxwxghp9iMcYYUwgndz3NE5GaOZZrichrJThmMnBpjuWGuKrTGmOMCUBOLmS1UdVj2QuqehRoW4JjbgKaikhj94RII4AVJdifh4gMEZHZx48fL43dGWOMwVmiqOS+LRYAEamNsye6EZFFwJdAcxFJFpFbVDUDmAisAr4D3lLVUqkZpaorVXVcjRo1SmN3xhhjcPaF/yzwhYgsdS9fA/zdyc5VdWQB7R9Shg/tGWOMKT4nJTzmi0gS0BvXQPQfVHWn3yMrBhEZAgyJjo4u71CMMSZoOLrZ1n1p6C1gOXAyZzXZQGKXnowxpvQ5uetpqIj8APwH+AzYB3zk57iMMcYECCdnFI8CXYDdqtoY6At87teojDHGBAwniSJdVQ/juvupkqquAeL9HJcxxpgA4eSup2MiUh1IBBaKyEEgw79hFY8NZhtjTOlzckYxDDgN3IWrIOBe8teBCgg2mG2MMaXPye2xp9xvs0TkA+CwFlRy1hhjTNAp8IxCRLqIyFoReVdE2ronH9oBHBCRgQV9zhhjTHDxdUbxInAfUAP4FLhSVTeISAtgEe55KQKJjVEYY0zp8zVGcZ6qrlbVt4FfVXUDgKp+XzahFZ2NURhjTOnzlSiycrw/k2edjVEYY0yI8HXpKU5ETuCq71TF/R73coTfIzPGGBMQfM1wF1aWgRhjjAlMxZuB2xhjTMgIqkRhM9wZY0zpC6pEYXc9GWNM6QuqRGGMMab0WaIwxhjjkyUKY4wxPlmiMMYY45MlCmOMMT4FVaKw22ONMab0BVWisNtjjTGm9AVVojDGGFP6LFEYY4zxyRKFMcYYnyxRGGOM8ckShTHGGJ8sURhjjPHJEoUxxhifgipR2AN3xhhT+oIqUdgDd8YYU/qCKlEYY4wpfZYojDHG+GSJwhhjjE+WKIwxxvhkicIYY4xPliiMMcb4ZInCGGOMT5YojDHG+GSJwhhjjE+WKIwxxvhkicIYY4xPQZUorCigMcaUvqBKFFYU0BhjSl9QJQpjjDGlzxKFMcYYnyxRGGOM8ckShTHGGJ8sURhjjPHJEoUxxhifLFEYY4zxyRKFMcYYnyxRGGOM8ckShTHGGJ8sURhjjPHJEoUxxhifLFEYY4zxyRKFMcYYnyxRGGOM8ckShTHGGJ8CPlGISEsRmSUiS0VkQnnHY4wxocaviUJEXhORgyKyI0/7QBHZJSJ7ROReX/tQ1e9U9TbgWqCDP+M1xpiK6O3db5OJ+m3//j6jmAsMzNkgImHAP4ErgRhgpIjEiEhrEXk/z+ti92eGAuuBf/s5XmOMqXA+/PFDAML8tP/z/LRfAFQ1UUSi8jR3Avao6o8AIrIYGKaqTwCDC9jPCmCFiHwAvOm/iI0xpmIKQzjPTycVfk0UBWgA/DfHcjLQuaCNRaQX8AfgfOBDH9uNA8a5F0+KyK5ixnch8FsxP1tRWZ9Dg/U5iM1lLsCFLLitJP1t5K2xPBKFeGkrMA+q6lpgbWE7VdXZwOxiR+UmIptVNaTGQqzPocH6HPz81d/yuOspGbg0x3JDYH85xGGMMcaB8kgUm4CmItJYRCoDI4AV5RCHMcYYB/x9e+wi4EuguYgki8gtqpoBTARWAd8Bb6nqt/6Mo4hKfPmqArI+hwbrc/DzS39F1X/33hpjjKn4Av7JbGOMMeUrZBNFYU+Hi8j5IrLEvf4rL8+DVDgO+vxnEdkpIttE5N8i4vVWuYrEaRUAEfmjiKiIVOg7ZJz0V0Sudf89fysiFf65JAf/rn8nImtEZIv73/ag8oizNBVU9SLHehGRGe6fyTYRaVeiA6pqyL1wPcC4F2gCVAa+AWLybHM7MMv9fgSwpLzjLoM+9waqut9PCIU+u7eLBBKBDUCH8o7bz3/HTYEtQC338sXlHXcZ9Hk2MMH9PgbYV95xl0K/ewDtgB0FrB8EfITrcYQuwFclOV6onlF4ng5X1XPAYmBYnm2GAfPc75cCfUXE2zMgFUWhfVbVNap62r24AdetyxWZk79ngEeBp4G0sgzOD5z091bgn6p6FEBVD5ZxjKXNSZ8VuMD9vgZBcDu+qiYCR3xsMgyYry4bgJoiUq+4xwvVROHt6fAGBW2jrju1jgN1yiQ6/3DS55xuwfUbSUVWaJ9FpC1wqaq+X5aB+YmTv+NmQDMR+VxENojIQCo2J31+CLhBRJJxVXe4s2xCK1dF/f/uU3k8mR0InDwdXqQnyCsAx/0RkRtwVert6deI/M9nn0WkEvC/wE1lFZCfOfk7Pg/X5adeuM4Y14lIK1U95ufY/MVJn0cCc1X1WRHpCrzh7nOW/8MrN6X6/RWqZxROng73bCMi5+E6ZfV1qhfoHD0RLyL9gPuBoap6toxi85fC+hwJtALWisg+XNdyV1TgAW2n/66Xq2q6qv4H2IUrcVRUTvp8C/AWgKp+CUTgqgEVzEq1AkaoJgonT4evAMa43/8R+FTdo0QVVKF9dl+GeRlXkqjo166hkD6r6nFVvVBVo1Q1Cte4zFBV3Vw+4ZaYk3/X7+G6aQERuRDXpagfyzTK0uWkzz8DfcE1ERquRHGoTKMseyuA0e67n7oAx1U1pbg7C8lLT6qaISLZT4eHAa+p6rci8giwWV1lzefgOkXdg+tMYkT5RVxyDvv8DFAdeNs9bv+zqg4tt6BLyGGfg4bD/q4C+ovITiATuFtVD5df1CXjsM9/AV4RkbtwXX65qYL/0pdd9aIXcKF77GUaEA6gqrNwjcUMAvYAp4H/KdHxKvjPyxhjjJ+F6qUnY4wxDlmiMMYY45MlCmOMMT5ZojDGGOOTJQpjjDE+WaIwQUVE7ndXRd0mIltFpLO7/VURifHD8U4W0J7pPn726153e3d3fFtFpIqIPONefkZEbhOR0T6OVV9ElpZ2H4wpjN0ea4KGuzzDc0AvVT3rfqCssqr6rQiciJxU1epFaJ+Fq5Ln6+7lE8BFQfAUvAlidkZhgkk94LfsL11V/S07SYjI2uzSHCJyi4jsdre9IiIvutvnumv4fyEiP4rIH93t1cU1P8fXIrJdRLxVoC2UiIwFrgUeFJGFIrICqAZ8JSLXichDIvJX97bRIvIvEfnGfdzLRCQqe/4BEQlzn4Vscp89jXe393L3a6mIfO8+jrjXdXT37RsR2SgikSKyTkTic8T4uYi0KU7/TPAKySezTdBajetLeDfwL1zzaXyWcwMRqQ9MxVXLPxX4FNccBtnqAd2AFrjKICzFVX58uKqecJ+lbBCRFYU83VtFRLbmWH5CVV8VkW7A+6q61B3PSVWNd79/KMf2C4EnVXWZiETg+qXu4hzrb8FVlqGjiJwPfC4iq93r2gKxuGr7fA5cISIbgSXAdaq6SUQuAM4Ar+IqivgnEWkGnK+q23z0y4QgO6MwQUNVTwLtgXG4avksEZGb8mzWCfhMVY+oajrwdp7176lqlqruBOq62wR4XES24UpADXKsK8gZVY3P8VritB8iEgk0UNVl7n6l5ZgnJFt/XLV8tgJf4SqBn13cb6OqJruro24FooDmQIqqbnLv84S7fP7bwGARCQduBuY6jdOEDjujMEFFVTOBtbgqwm7HVdhxbo5NCpt8KudYQfa2o4CLgPaqmi6uSrMRpRFvAZxMkCXAnaq6KlejSC9y9yET1/9zwUuZaVU9LSKf4Jro5lpc5eWNycXOKEzQEJHmIpKzZHY88FOezTYCPUWklrjKx1/tYNc1gIPuJNEb8Otc4qp6AkgWkd+DZ/72qnk2WwVMcJ8JICLNRKSaj91+D9QXkY7u7SPd/QfX5acZwCZVrcil9I2f2BmFCSbVgX+ISE0gA1flzHE5N1DVX0TkcVyXa/YDO3HNXujLQmCliGzGdSnnewex5B2j+FhV73XWDQBuBF52V0FNB64Bck608yquS0pfuwerDwG/L2hnqnpORK7D9fOpgmt8oh9wUlWT3HdfvV6E+EwIsdtjTcgRkeqqetL9G/UyXKWpl5V3XOXFPcC/FmgR5LO+mWKyS08mFD3k/m1/B/AfXJP5hCT3A35fAfdbkjAFsTMKY4wxPtkZhTHGGJ8sURhjjPHJEoUxxhifLFEYY4zxyRKFMcYYnyxRGGOM8en/AMaTjNvefeiSAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot confusion matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot normalized confusion matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Plot ROC curve\")\n",
    "y_predict = makeRoc(X_test, labels, y_test, model)\n",
    "plt.show()\n",
    "\n",
    "print(\"Plot confusion matrix\")\n",
    "y_test_proba = y_test.argmax(axis=1)\n",
    "y_predict_proba = y_predict.argmax(axis=1)\n",
    "\n",
    "# Compute non-normalized confusion matrix\n",
    "cnf_matrix = confusion_matrix(y_test_proba, y_predict_proba)\n",
    "np.set_printoptions(precision=2)\n",
    "# Plot non-normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=[l.replace('j_','') for l in labels],\n",
    "                          title='Confusion matrix')\n",
    "plt.show()\n",
    "\n",
    "print(\"Plot normalized confusion matrix\")\n",
    "# Plot normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=[l.replace('j_','') for l in labels], normalize=True,\n",
    "                          title='Normalized confusion matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}