DNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:469: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:470: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:471: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:472: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:473: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:476: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n",
      "/home/anrunw/anaconda3/envs/tensorflow/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6\n",
      "  return f(*args, **kwds)\n"
     ]
    },
    {
     "ename": "ImportError",
     "evalue": "Keras requires TensorFlow 2.2 or higher. Install TensorFlow via `pip install tensorflow`",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m~/anaconda3/envs/tensorflow/lib/python3.6/site-packages/keras/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m     \u001b[0;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexperimental\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpreprocessing\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mRandomRotation\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      4\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mImportError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'tensorflow.keras'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[0;31mImportError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-0d1135a6788e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtensorflow\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mInput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mDense\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodels\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mModel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptimizers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mAdam\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/tensorflow/lib/python3.6/site-packages/keras/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mImportError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     raise ImportError(\n\u001b[0;32m----> 6\u001b[0;31m         \u001b[0;34m'Keras requires TensorFlow 2.2 or higher. '\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m         'Install TensorFlow via `pip install tensorflow`')\n\u001b[1;32m      8\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mImportError\u001b[0m: Keras requires TensorFlow 2.2 or higher. Install TensorFlow via `pip install tensorflow`"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import keras\n",
    "from keras.layers import Input, Dense\n",
    "from keras.models import Model\n",
    "from keras.optimizers import Adam\n",
    "from sklearn.model_selection import train_test_split\n",
    "import os\n",
    "import h5py\n",
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'pd' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-5-e70138d57952>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mdg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread_hdf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'train.h5'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'table'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'pd' is not defined"
     ]
    }
   ],
   "source": [
    "dg = pd.read_hdf('train.h5', 'table')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'dg' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-6-41d808eb4aa0>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0my_train\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'is_signal_new'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto_numpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'dg' is not defined"
     ]
    }
   ],
   "source": [
    "y_train = dg['is_signal_new'].to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'dg' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-2-f37d1f089ca9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mdg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdrop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'ttv'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthE'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPX'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPY'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPZ'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'dg' is not defined"
     ]
    }
   ],
   "source": [
    "dg = dg.drop(['ttv', 'truthE', 'truthPX', 'truthPY', 'truthPZ'], axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'dg' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-3-65cd73e0d787>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mdg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdrop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'is_signal_new'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'dg' is not defined"
     ]
    }
   ],
   "source": [
    "dg = dg.drop(['is_signal_new'], axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'dg' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-bbd351b232b3>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mX_train\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto_numpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'dg' is not defined"
     ]
    }
   ],
   "source": [
    "X_train = dg.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_6 (InputLayer)         [(None, 800)]             0         \n",
      "_________________________________________________________________\n",
      "fc1_relu (Dense)             (None, 64)                51264     \n",
      "_________________________________________________________________\n",
      "fc2 (Dense)                  (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "fc3 (Dense)                  (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "fc4 (Dense)                  (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "output_sigmoid (Dense)       (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 55,489\n",
      "Trainable params: 55,489\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "Inputs = Input(shape=(800))\n",
    "x = Dense(64, activation='relu', kernel_initializer='lecun_uniform', name='fc1_relu')(Inputs)\n",
    "y = Dense(32, activation='relu', kernel_initializer='lecun_uniform', name = 'fc2')(x)\n",
    "z = Dense(32, activation='relu', kernel_initializer='lecun_uniform', name = 'fc3')(y)\n",
    "a = Dense(32, activation='relu', kernel_initializer='lecun_uniform', name='fc4')(z)\n",
    "predictions = Dense(1, activation='sigmoid', kernel_initializer='lecun_uniform', name = 'output_sigmoid')(a)\n",
    "model = Model(inputs=Inputs, outputs=predictions)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "adam = Adam(lr=0.0001)\n",
    "model.compile(optimizer=adam, loss='binary_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 474.07114 , -250.34703 , -223.65196 , ...,    0.      ,\n",
       "           0.      ,    0.      ],\n",
       "       [ 150.50453 ,  120.06239 ,   76.852005, ...,    0.      ,\n",
       "           0.      ,    0.      ],\n",
       "       [ 251.64539 ,   10.427651, -147.57375 , ...,    0.      ,\n",
       "           0.      ,    0.      ],\n",
       "       ...,\n",
       "       [  61.417538,   42.90129 ,   43.947723, ...,    0.      ,\n",
       "           0.      ,    0.      ],\n",
       "       [ 261.2153  ,   12.780115, -132.6992  , ...,    0.      ,\n",
       "           0.      ,    0.      ],\n",
       "       [ 221.00002 ,  -32.45829 , -127.960335, ...,    0.      ,\n",
       "           0.      ,    0.      ]], dtype=float32)"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "887/887 [==============================] - 9s 10ms/step - loss: 0.5731 - accuracy: 0.6943 - val_loss: 0.5410 - val_accuracy: 0.7147\n",
      "Epoch 2/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.5309 - accuracy: 0.7199 - val_loss: 0.5252 - val_accuracy: 0.7244\n",
      "Epoch 3/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.5139 - accuracy: 0.7320 - val_loss: 0.5060 - val_accuracy: 0.7386\n",
      "Epoch 4/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4952 - accuracy: 0.7473 - val_loss: 0.4897 - val_accuracy: 0.7523\n",
      "Epoch 5/100\n",
      "887/887 [==============================] - 7s 7ms/step - loss: 0.4778 - accuracy: 0.7636 - val_loss: 0.4740 - val_accuracy: 0.7698\n",
      "Epoch 6/100\n",
      "887/887 [==============================] - 6s 7ms/step - loss: 0.4626 - accuracy: 0.7780 - val_loss: 0.4630 - val_accuracy: 0.7790\n",
      "Epoch 7/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4498 - accuracy: 0.7899 - val_loss: 0.4465 - val_accuracy: 0.7951\n",
      "Epoch 8/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4387 - accuracy: 0.7994 - val_loss: 0.4396 - val_accuracy: 0.8002\n",
      "Epoch 9/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4293 - accuracy: 0.8064 - val_loss: 0.4322 - val_accuracy: 0.8044\n",
      "Epoch 10/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4210 - accuracy: 0.8122 - val_loss: 0.4223 - val_accuracy: 0.8135\n",
      "Epoch 11/100\n",
      "887/887 [==============================] - 7s 7ms/step - loss: 0.4132 - accuracy: 0.8183 - val_loss: 0.4237 - val_accuracy: 0.8097\n",
      "Epoch 12/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4068 - accuracy: 0.8221 - val_loss: 0.4101 - val_accuracy: 0.8200\n",
      "Epoch 13/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.4002 - accuracy: 0.8262 - val_loss: 0.4044 - val_accuracy: 0.8229\n",
      "Epoch 14/100\n",
      "887/887 [==============================] - 7s 7ms/step - loss: 0.3933 - accuracy: 0.8301 - val_loss: 0.3954 - val_accuracy: 0.8286\n",
      "Epoch 15/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3874 - accuracy: 0.8330 - val_loss: 0.3890 - val_accuracy: 0.8324\n",
      "Epoch 16/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3817 - accuracy: 0.8365 - val_loss: 0.3863 - val_accuracy: 0.8329\n",
      "Epoch 17/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3763 - accuracy: 0.8394 - val_loss: 0.3792 - val_accuracy: 0.8384\n",
      "Epoch 18/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3711 - accuracy: 0.8423 - val_loss: 0.3761 - val_accuracy: 0.8389\n",
      "Epoch 19/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3668 - accuracy: 0.8443 - val_loss: 0.3706 - val_accuracy: 0.8423\n",
      "Epoch 20/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3621 - accuracy: 0.8467 - val_loss: 0.3659 - val_accuracy: 0.8450\n",
      "Epoch 21/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3587 - accuracy: 0.8484 - val_loss: 0.3649 - val_accuracy: 0.8451\n",
      "Epoch 22/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3549 - accuracy: 0.8504 - val_loss: 0.3796 - val_accuracy: 0.8355\n",
      "Epoch 23/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3509 - accuracy: 0.8521 - val_loss: 0.3619 - val_accuracy: 0.8457\n",
      "Epoch 24/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3475 - accuracy: 0.8539 - val_loss: 0.3521 - val_accuracy: 0.8518\n",
      "Epoch 25/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3441 - accuracy: 0.8557 - val_loss: 0.3607 - val_accuracy: 0.8455\n",
      "Epoch 26/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3410 - accuracy: 0.8576 - val_loss: 0.3477 - val_accuracy: 0.8533\n",
      "Epoch 27/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3384 - accuracy: 0.8587 - val_loss: 0.3421 - val_accuracy: 0.8566\n",
      "Epoch 28/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3357 - accuracy: 0.8599 - val_loss: 0.3410 - val_accuracy: 0.8571\n",
      "Epoch 29/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3326 - accuracy: 0.8617 - val_loss: 0.3419 - val_accuracy: 0.8565\n",
      "Epoch 30/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3304 - accuracy: 0.8626 - val_loss: 0.3395 - val_accuracy: 0.8567\n",
      "Epoch 31/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3286 - accuracy: 0.8635 - val_loss: 0.3365 - val_accuracy: 0.8584\n",
      "Epoch 32/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3262 - accuracy: 0.8647 - val_loss: 0.3322 - val_accuracy: 0.8610\n",
      "Epoch 33/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3246 - accuracy: 0.8651 - val_loss: 0.3322 - val_accuracy: 0.8615\n",
      "Epoch 34/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3225 - accuracy: 0.8664 - val_loss: 0.3302 - val_accuracy: 0.8622\n",
      "Epoch 35/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3211 - accuracy: 0.8668 - val_loss: 0.3271 - val_accuracy: 0.8637\n",
      "Epoch 36/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3187 - accuracy: 0.8682 - val_loss: 0.3330 - val_accuracy: 0.8593\n",
      "Epoch 37/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3176 - accuracy: 0.8689 - val_loss: 0.3395 - val_accuracy: 0.8571\n",
      "Epoch 38/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3160 - accuracy: 0.8696 - val_loss: 0.3201 - val_accuracy: 0.8674\n",
      "Epoch 39/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3150 - accuracy: 0.8702 - val_loss: 0.3253 - val_accuracy: 0.8640\n",
      "Epoch 40/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3137 - accuracy: 0.8708 - val_loss: 0.3217 - val_accuracy: 0.8667\n",
      "Epoch 41/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3115 - accuracy: 0.8721 - val_loss: 0.3187 - val_accuracy: 0.8684\n",
      "Epoch 42/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3107 - accuracy: 0.8722 - val_loss: 0.3228 - val_accuracy: 0.8662\n",
      "Epoch 43/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3096 - accuracy: 0.8726 - val_loss: 0.3220 - val_accuracy: 0.8656\n",
      "Epoch 44/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3089 - accuracy: 0.8732 - val_loss: 0.3296 - val_accuracy: 0.8628\n",
      "Epoch 45/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3076 - accuracy: 0.8739 - val_loss: 0.3205 - val_accuracy: 0.8664\n",
      "Epoch 46/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3069 - accuracy: 0.8740 - val_loss: 0.3131 - val_accuracy: 0.8719\n",
      "Epoch 47/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3058 - accuracy: 0.8748 - val_loss: 0.3112 - val_accuracy: 0.8723\n",
      "Epoch 48/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.3045 - accuracy: 0.8755 - val_loss: 0.3173 - val_accuracy: 0.8685\n",
      "Epoch 49/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3044 - accuracy: 0.8756 - val_loss: 0.3131 - val_accuracy: 0.8712\n",
      "Epoch 50/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3033 - accuracy: 0.8760 - val_loss: 0.3089 - val_accuracy: 0.8732\n",
      "Epoch 51/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3021 - accuracy: 0.8765 - val_loss: 0.3098 - val_accuracy: 0.8727\n",
      "Epoch 52/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.3010 - accuracy: 0.8771 - val_loss: 0.3215 - val_accuracy: 0.8662\n",
      "Epoch 53/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2999 - accuracy: 0.8775 - val_loss: 0.3058 - val_accuracy: 0.8748\n",
      "Epoch 54/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2996 - accuracy: 0.8777 - val_loss: 0.3067 - val_accuracy: 0.8753\n",
      "Epoch 55/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2980 - accuracy: 0.8787 - val_loss: 0.3135 - val_accuracy: 0.8707\n",
      "Epoch 56/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2983 - accuracy: 0.8783 - val_loss: 0.3072 - val_accuracy: 0.8743\n",
      "Epoch 57/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2969 - accuracy: 0.8791 - val_loss: 0.3039 - val_accuracy: 0.8761\n",
      "Epoch 58/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2963 - accuracy: 0.8793 - val_loss: 0.3106 - val_accuracy: 0.8737\n",
      "Epoch 59/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2970 - accuracy: 0.8787 - val_loss: 0.3097 - val_accuracy: 0.8728\n",
      "Epoch 60/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2952 - accuracy: 0.8798 - val_loss: 0.3015 - val_accuracy: 0.8772\n",
      "Epoch 61/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2946 - accuracy: 0.8803 - val_loss: 0.3023 - val_accuracy: 0.8771\n",
      "Epoch 62/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2938 - accuracy: 0.8804 - val_loss: 0.3047 - val_accuracy: 0.8749\n",
      "Epoch 63/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2939 - accuracy: 0.8802 - val_loss: 0.3068 - val_accuracy: 0.8741\n",
      "Epoch 64/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2928 - accuracy: 0.8811 - val_loss: 0.2980 - val_accuracy: 0.8791\n",
      "Epoch 65/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2923 - accuracy: 0.8812 - val_loss: 0.3034 - val_accuracy: 0.8755\n",
      "Epoch 66/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2915 - accuracy: 0.8816 - val_loss: 0.3007 - val_accuracy: 0.8781\n",
      "Epoch 67/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2924 - accuracy: 0.8810 - val_loss: 0.3012 - val_accuracy: 0.8775\n",
      "Epoch 68/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2908 - accuracy: 0.8819 - val_loss: 0.2994 - val_accuracy: 0.8786\n",
      "Epoch 69/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2895 - accuracy: 0.8827 - val_loss: 0.2963 - val_accuracy: 0.8798\n",
      "Epoch 70/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2904 - accuracy: 0.8821 - val_loss: 0.3008 - val_accuracy: 0.8774\n",
      "Epoch 71/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2893 - accuracy: 0.8829 - val_loss: 0.2995 - val_accuracy: 0.8770\n",
      "Epoch 72/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2888 - accuracy: 0.8827 - val_loss: 0.2982 - val_accuracy: 0.8782\n",
      "Epoch 73/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2883 - accuracy: 0.8832 - val_loss: 0.2965 - val_accuracy: 0.8800\n",
      "Epoch 74/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2876 - accuracy: 0.8836 - val_loss: 0.3000 - val_accuracy: 0.8780\n",
      "Epoch 75/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2884 - accuracy: 0.8829 - val_loss: 0.3126 - val_accuracy: 0.8707\n",
      "Epoch 76/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2867 - accuracy: 0.8839 - val_loss: 0.2939 - val_accuracy: 0.8804\n",
      "Epoch 77/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2864 - accuracy: 0.8840 - val_loss: 0.3029 - val_accuracy: 0.8761\n",
      "Epoch 78/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2870 - accuracy: 0.8837 - val_loss: 0.3016 - val_accuracy: 0.8767\n",
      "Epoch 79/100\n",
      "887/887 [==============================] - 8s 8ms/step - loss: 0.2862 - accuracy: 0.8841 - val_loss: 0.2962 - val_accuracy: 0.8796\n",
      "Epoch 80/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2855 - accuracy: 0.8844 - val_loss: 0.2956 - val_accuracy: 0.8794\n",
      "Epoch 81/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2855 - accuracy: 0.8846 - val_loss: 0.2942 - val_accuracy: 0.8804\n",
      "Epoch 82/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2857 - accuracy: 0.8841 - val_loss: 0.2927 - val_accuracy: 0.8822\n",
      "Epoch 83/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2847 - accuracy: 0.8848 - val_loss: 0.2944 - val_accuracy: 0.8813\n",
      "Epoch 84/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2839 - accuracy: 0.8852 - val_loss: 0.2934 - val_accuracy: 0.8812\n",
      "Epoch 85/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2851 - accuracy: 0.8845 - val_loss: 0.3014 - val_accuracy: 0.8765\n",
      "Epoch 86/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2830 - accuracy: 0.8857 - val_loss: 0.2942 - val_accuracy: 0.8804\n",
      "Epoch 87/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2833 - accuracy: 0.8855 - val_loss: 0.2952 - val_accuracy: 0.8799\n",
      "Epoch 88/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2842 - accuracy: 0.8848 - val_loss: 0.2956 - val_accuracy: 0.8806\n",
      "Epoch 89/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2835 - accuracy: 0.8852 - val_loss: 0.2905 - val_accuracy: 0.8827\n",
      "Epoch 90/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2820 - accuracy: 0.8859 - val_loss: 0.2923 - val_accuracy: 0.8820\n",
      "Epoch 91/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2824 - accuracy: 0.8857 - val_loss: 0.2911 - val_accuracy: 0.8817\n",
      "Epoch 92/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2818 - accuracy: 0.8859 - val_loss: 0.2949 - val_accuracy: 0.8802\n",
      "Epoch 93/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2828 - accuracy: 0.8855 - val_loss: 0.2978 - val_accuracy: 0.8784\n",
      "Epoch 94/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2811 - accuracy: 0.8866 - val_loss: 0.2917 - val_accuracy: 0.8817\n",
      "Epoch 95/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2814 - accuracy: 0.8864 - val_loss: 0.2952 - val_accuracy: 0.8803\n",
      "Epoch 96/100\n",
      "887/887 [==============================] - 8s 9ms/step - loss: 0.2813 - accuracy: 0.8866 - val_loss: 0.2918 - val_accuracy: 0.8821\n",
      "Epoch 97/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2806 - accuracy: 0.8866 - val_loss: 0.2933 - val_accuracy: 0.8809\n",
      "Epoch 98/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2805 - accuracy: 0.8868 - val_loss: 0.2985 - val_accuracy: 0.8781\n",
      "Epoch 99/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2802 - accuracy: 0.8868 - val_loss: 0.2970 - val_accuracy: 0.8786\n",
      "Epoch 100/100\n",
      "887/887 [==============================] - 7s 8ms/step - loss: 0.2796 - accuracy: 0.8869 - val_loss: 0.2928 - val_accuracy: 0.8814\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, y_train, batch_size = 1024, epochs = 100, \n",
    "                    validation_split = 0.25, shuffle = True, callbacks = None,\n",
    "                    use_multiprocessing=True, workers=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "def learningCurve(history):\n",
    "    plt.figure(figsize=(10,8))\n",
    "    plt.plot(history.history['loss'], linewidth=1)\n",
    "    plt.plot(history.history['val_loss'], linewidth=1)\n",
    "    plt.title('Model Loss over Epochs')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.xlabel('Epoch')\n",
    "    plt.legend(['training sample loss','validation sample loss'])\n",
    "    #plt.savefig('Learning_curve.pdf')\n",
    "    plt.show()\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learningCurve(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = ['top']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "cannot create a storer if the object is not existing nor a value are passed",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-58-620b54701e14>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mdg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread_hdf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'test.h5'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'table'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0my_test\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'is_signal_new'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto_numpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mdg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdrop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'ttv'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthE'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPX'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPY'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'truthPZ'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/tfgpu/lib/python3.6/site-packages/pandas/io/pytables.py\u001b[0m in \u001b[0;36mread_hdf\u001b[0;34m(path_or_buf, key, mode, errors, where, start, stop, columns, iterator, chunksize, **kwargs)\u001b[0m\n\u001b[1;32m    413\u001b[0m             \u001b[0miterator\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0miterator\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    414\u001b[0m             \u001b[0mchunksize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchunksize\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 415\u001b[0;31m             \u001b[0mauto_close\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mauto_close\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    416\u001b[0m         )\n\u001b[1;32m    417\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mValueError\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mTypeError\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/tfgpu/lib/python3.6/site-packages/pandas/io/pytables.py\u001b[0m in \u001b[0;36mselect\u001b[0;34m(self, key, where, start, stop, columns, iterator, chunksize, auto_close)\u001b[0m\n\u001b[1;32m    820\u001b[0m         \u001b[0;31m# create the storer and axes\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    821\u001b[0m         \u001b[0mwhere\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_ensure_term\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscope_level\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 822\u001b[0;31m         \u001b[0ms\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_create_storer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgroup\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    823\u001b[0m         \u001b[0ms\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minfer_axes\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    824\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/tfgpu/lib/python3.6/site-packages/pandas/io/pytables.py\u001b[0m in \u001b[0;36m_create_storer\u001b[0;34m(self, group, format, value, encoding, errors)\u001b[0m\n\u001b[1;32m   1612\u001b[0m                 \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1613\u001b[0m                     raise TypeError(\n\u001b[0;32m-> 1614\u001b[0;31m                         \u001b[0;34m\"cannot create a storer if the object is not existing \"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1615\u001b[0m                         \u001b[0;34m\"nor a value are passed\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1616\u001b[0m                     )\n",
      "\u001b[0;31mTypeError\u001b[0m: cannot create a storer if the object is not existing nor a value are passed"
     ]
    }
   ],
   "source": [
    "dg = pd.read_hdf('test.h5', 'table')\n",
    "\n",
    "y_test = dg['is_signal_new'].to_numpy()\n",
    "\n",
    "dg = dg.drop(['ttv', 'truthE', 'truthPX', 'truthPY', 'truthPZ'], axis = 1)\n",
    "\n",
    "dg = dg.drop(['is_signal_new'], axis = 1)\n",
    "\n",
    "X_test = dg.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "def makeRoc(features_val, labels, labels_val, model):\n",
    "    print('in makeRoc()')\n",
    "    if 'j_index' in labels: labels.remove('j_index')\n",
    "\n",
    "    predict_test = model.predict(features_val)\n",
    "\n",
    "    df = pd.DataFrame()\n",
    "    \n",
    "    fpr = {}\n",
    "    tpr = {}\n",
    "    auc1 = {}\n",
    "    \n",
    "    plt.figure()       \n",
    "    for i, label in enumerate(labels):\n",
    "        df[label] = labels_val[:,i]\n",
    "        df[label + '_pred'] = predict_test[:,i]\n",
    "        \n",
    "        fpr[label], tpr[label], threshold = roc_curve(df[label],df[label+'_pred'])\n",
    "\n",
    "        auc1[label] = auc(fpr[label], tpr[label])\n",
    "            \n",
    "        plt.plot(tpr[label],fpr[label],label='%s tagger, AUC = %.1f%%'%(label.replace('j_',''),auc1[label]*100.))\n",
    "    plt.semilogy()\n",
    "    plt.xlabel(\"Signal Efficiency\")\n",
    "    plt.ylabel(\"Background Efficiency\")\n",
    "    plt.ylim(0.001,1)\n",
    "    plt.grid(True)\n",
    "    plt.legend(loc='upper left')\n",
    "    plt.figtext(0.25, 0.90,'hls4ml',fontweight='bold', wrap=True, horizontalalignment='right', fontsize=14)\n",
    "    #plt.figtext(0.35, 0.90,'preliminary', style='italic', wrap=True, horizontalalignment='center', fontsize=14) \n",
    "    plt.savefig(\"mygraph.png\")\n",
    "#     plt.savefig('%s/ROC.pdf'%(options.outputDir))\n",
    "#     plt.savefig('%s/ROC.pdf' %(outputDir))\n",
    "    return predict_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "ename": "IndexError",
     "evalue": "too many indices for array: array is 1-dimensional, but 2 were indexed",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-40-84baff8a63b6>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0my_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmakeRoc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_test\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_test\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabels\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m<ipython-input-38-ee94d122bac8>\u001b[0m in \u001b[0;36mmakeRoc\u001b[0;34m(features_val, labels_val, labels, model, outputDir, outputSuffix)\u001b[0m\n\u001b[1;32m      8\u001b[0m     \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfigure\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfigsize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m8\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabel\u001b[0m \u001b[0;32min\u001b[0m \u001b[0menumerate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlabels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m         \u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlabels_val\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m         \u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m'_pred'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlabels_pred\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m         \u001b[0mfpr\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtpr\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mthreshold\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mroc_curve\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlabel\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;34m'_pred'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mIndexError\u001b[0m: too many indices for array: array is 1-dimensional, but 2 were indexed"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 720x576 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred = makeRoc(x_test, labels, y_test, model)"
   ]
  },
  {
   "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": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}