1. 1DCNN + LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## References:\n",
    "Part of code taken from: https://github.com/keras-team/keras/blob/master/examples/imdb_cnn_lstm.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing import sequence\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation\n",
    "from keras.layers import Embedding\n",
    "from keras.layers import LSTM, BatchNormalization\n",
    "from keras.layers import Conv1D, MaxPooling1D\n",
    "from keras.datasets import imdb\n",
    "from utils import *\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get data and define features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Get raw documents:\n",
    "train, test, labels = getData()\n",
    "train, _X, labels, _Y = train_test_split(train, labels, test_size=0.20, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Convert chars to integer:\n",
    "#Use a charcter tokenizer from utils\n",
    "char_tokenizer = Char_tokenizer(train+test+_X)\n",
    "X_train = char_tokenizer.transform(train)\n",
    "X_test = char_tokenizer.transform(test)\n",
    "__X = char_tokenizer.transform(_X)\n",
    "#and one hot encode y:\n",
    "num_of_classes = 12\n",
    "y_train = get1hotY(labels, num_of_classes)\n",
    "__Y = get1hotY(_Y, num_of_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((26010, 452), (3000, 452), (26010, 12), (6503, 452), (6503, 12))"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, X_test.shape, y_train.shape, __X.shape, __Y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_features = 27\n",
    "# We are trying to learn the embedding on char level, the intuition behind embedding_size = number of classes\n",
    "# is that I am trying to create embedding for classes.\n",
    "# However I did some empericals expeiments and it turns out that increaing the number of embeddings (50,100)\n",
    "# doesn't help in accuracy  \n",
    "embedding_size = 13\n",
    "max_len = getMaxLen(train+test)\n",
    "num_classes = y_train.shape[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Lets model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Original model from keras example with embedding layer at the start"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "def originalModel(max_features, embedding_size, maxlen, num_classes):\n",
    "    \n",
    "    kernel_size = 5\n",
    "    filters = 64\n",
    "    pool_size = 4\n",
    "    lstm_output_size = 200\n",
    "    \n",
    "    model = Sequential()\n",
    "    model.add(Embedding(max_features, embedding_size, input_length=maxlen))\n",
    "    model.add(BatchNormalization())\n",
    "    #model.add(Dropout(0.20))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 4, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 5, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 6, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 8, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(LSTM(lstm_output_size))\n",
    "    model.add(Dense(100))\n",
    "    model.add(Dense(50))\n",
    "    model.add(Dense(num_classes))\n",
    "    model.add(Activation('sigmoid'))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import *\n",
    "def f1_score(y_true, y_pred):\n",
    "\n",
    "    # Count positive samples.\n",
    "    c1 = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
    "    c2 = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
    "    c3 = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
    "\n",
    "    # If there are no true samples, fix the F1 score at 0.\n",
    "    if c3 == 0:\n",
    "        return 0\n",
    "\n",
    "    # How many selected items are relevant?\n",
    "    precision = c1 / c2\n",
    "\n",
    "    # How many relevant items are selected?\n",
    "    recall = c1 / c3\n",
    "\n",
    "    # Calculate f1_score\n",
    "    f1_score = 2 * (precision * recall) / (precision + recall)\n",
    "    return f1_score "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 23409 samples, validate on 2601 samples\n",
      "Epoch 1/30\n",
      "23409/23409 [==============================] - 59s 3ms/step - loss: 2.3366 - acc: 0.1587 - f1_score: 0.2084 - val_loss: 2.3234 - val_acc: 0.1684 - val_f1_score: 0.2120\n",
      "Epoch 2/30\n",
      "23409/23409 [==============================] - 59s 3ms/step - loss: 2.1766 - acc: 0.2085 - f1_score: 0.2630 - val_loss: 2.0695 - val_acc: 0.2188 - val_f1_score: 0.2928\n",
      "Epoch 3/30\n",
      "23409/23409 [==============================] - 59s 3ms/step - loss: 1.9817 - acc: 0.2495 - f1_score: 0.3219 - val_loss: 1.9592 - val_acc: 0.2464 - val_f1_score: 0.3228\n",
      "Epoch 4/30\n",
      "23409/23409 [==============================] - 58s 2ms/step - loss: 1.7707 - acc: 0.3379 - f1_score: 0.3630 - val_loss: 1.6253 - val_acc: 0.4195 - val_f1_score: 0.3421\n",
      "Epoch 5/30\n",
      "23409/23409 [==============================] - 58s 2ms/step - loss: 1.5000 - acc: 0.4544 - f1_score: 0.3471 - val_loss: 1.4292 - val_acc: 0.4764 - val_f1_score: 0.3399\n",
      "Epoch 6/30\n",
      "23409/23409 [==============================] - 56s 2ms/step - loss: 1.3355 - acc: 0.5160 - f1_score: 0.3462 - val_loss: 1.2712 - val_acc: 0.5371 - val_f1_score: 0.4249\n",
      "Epoch 7/30\n",
      "23409/23409 [==============================] - 57s 2ms/step - loss: 1.1933 - acc: 0.5690 - f1_score: 0.3761 - val_loss: 1.1387 - val_acc: 0.5921 - val_f1_score: 0.3305\n",
      "Epoch 8/30\n",
      "23409/23409 [==============================] - 57s 2ms/step - loss: 1.1234 - acc: 0.5912 - f1_score: 0.3713 - val_loss: 1.1323 - val_acc: 0.5944 - val_f1_score: 0.3703\n",
      "Epoch 9/30\n",
      "23409/23409 [==============================] - 56s 2ms/step - loss: 1.0166 - acc: 0.6383 - f1_score: 0.3611 - val_loss: 1.0690 - val_acc: 0.6271 - val_f1_score: 0.2531\n",
      "Epoch 10/30\n",
      "23409/23409 [==============================] - 57s 2ms/step - loss: 0.9548 - acc: 0.6632 - f1_score: 0.3543 - val_loss: 1.0433 - val_acc: 0.6378 - val_f1_score: 0.3675\n",
      "Epoch 11/30\n",
      "23409/23409 [==============================] - 56s 2ms/step - loss: 0.8819 - acc: 0.6912 - f1_score: 0.3651 - val_loss: 1.0069 - val_acc: 0.6524 - val_f1_score: 0.3837\n",
      "Epoch 12/30\n",
      "23409/23409 [==============================] - 56s 2ms/step - loss: 0.8128 - acc: 0.7146 - f1_score: 0.3510 - val_loss: 1.0473 - val_acc: 0.6724 - val_f1_score: 0.4024\n",
      "Epoch 13/30\n",
      "23409/23409 [==============================] - 58s 2ms/step - loss: 0.7542 - acc: 0.7375 - f1_score: 0.4144 - val_loss: 0.9693 - val_acc: 0.6747 - val_f1_score: 0.3764\n",
      "Epoch 14/30\n",
      "23409/23409 [==============================] - 56s 2ms/step - loss: 0.7090 - acc: 0.7541 - f1_score: 0.3832 - val_loss: 0.9639 - val_acc: 0.6940 - val_f1_score: 0.4179\n",
      "Epoch 15/30\n",
      "23409/23409 [==============================] - 61s 3ms/step - loss: 0.6484 - acc: 0.7774 - f1_score: 0.4207 - val_loss: 1.0478 - val_acc: 0.6847 - val_f1_score: 0.3662\n",
      "Epoch 16/30\n",
      "23409/23409 [==============================] - 64s 3ms/step - loss: 0.6185 - acc: 0.7843 - f1_score: 0.4083 - val_loss: 0.9894 - val_acc: 0.6951 - val_f1_score: 0.3751\n",
      "Epoch 17/30\n",
      "23409/23409 [==============================] - 61s 3ms/step - loss: 0.5508 - acc: 0.8108 - f1_score: 0.4294 - val_loss: 1.0190 - val_acc: 0.6859 - val_f1_score: 0.4640\n",
      "Epoch 18/30\n",
      "23409/23409 [==============================] - 63s 3ms/step - loss: 0.5035 - acc: 0.8245 - f1_score: 0.4549 - val_loss: 1.0283 - val_acc: 0.6947 - val_f1_score: 0.3932\n",
      "Epoch 19/30\n",
      "23409/23409 [==============================] - 60s 3ms/step - loss: 0.4650 - acc: 0.8410 - f1_score: 0.5035 - val_loss: 1.1150 - val_acc: 0.6874 - val_f1_score: 0.4724\n",
      "Epoch 20/30\n",
      "23409/23409 [==============================] - 61s 3ms/step - loss: 0.4229 - acc: 0.8550 - f1_score: 0.5374 - val_loss: 1.0838 - val_acc: 0.7024 - val_f1_score: 0.4598\n",
      "Epoch 21/30\n",
      "23409/23409 [==============================] - 59s 3ms/step - loss: 0.3809 - acc: 0.8693 - f1_score: 0.5302 - val_loss: 1.0981 - val_acc: 0.7017 - val_f1_score: 0.4678\n",
      "Epoch 22/30\n",
      "23409/23409 [==============================] - 62s 3ms/step - loss: 0.3463 - acc: 0.8806 - f1_score: 0.5162 - val_loss: 1.1949 - val_acc: 0.6867 - val_f1_score: 0.5048\n",
      "Epoch 23/30\n",
      "23409/23409 [==============================] - 62s 3ms/step - loss: 0.3634 - acc: 0.8748 - f1_score: 0.5803 - val_loss: 1.1891 - val_acc: 0.6886 - val_f1_score: 0.5184\n",
      "Epoch 24/30\n",
      "23409/23409 [==============================] - 62s 3ms/step - loss: 0.2970 - acc: 0.8960 - f1_score: 0.5457 - val_loss: 1.3541 - val_acc: 0.6797 - val_f1_score: 0.4839\n",
      "Epoch 25/30\n",
      "23409/23409 [==============================] - 62s 3ms/step - loss: 0.2763 - acc: 0.9030 - f1_score: 0.5707 - val_loss: 1.3205 - val_acc: 0.6974 - val_f1_score: 0.5115\n",
      "Epoch 26/30\n",
      "23409/23409 [==============================] - 65s 3ms/step - loss: 0.2639 - acc: 0.9079 - f1_score: 0.6015 - val_loss: 1.4089 - val_acc: 0.6963 - val_f1_score: 0.5282\n",
      "Epoch 27/30\n",
      "23409/23409 [==============================] - 65s 3ms/step - loss: 0.2138 - acc: 0.9269 - f1_score: 0.6037 - val_loss: 1.3684 - val_acc: 0.6947 - val_f1_score: 0.5054\n",
      "Epoch 28/30\n",
      "23409/23409 [==============================] - 65s 3ms/step - loss: 0.2038 - acc: 0.9288 - f1_score: 0.6104 - val_loss: 1.4803 - val_acc: 0.7063 - val_f1_score: 0.4787\n",
      "Epoch 29/30\n",
      "23409/23409 [==============================] - 64s 3ms/step - loss: 0.1770 - acc: 0.9387 - f1_score: 0.5994 - val_loss: 1.4918 - val_acc: 0.7051 - val_f1_score: 0.5156\n",
      "Epoch 30/30\n",
      "23409/23409 [==============================] - 65s 3ms/step - loss: 0.1541 - acc: 0.9463 - f1_score: 0.6439 - val_loss: 1.6772 - val_acc: 0.7001 - val_f1_score: 0.5011\n"
     ]
    }
   ],
   "source": [
    "model = originalModel(max_features, embedding_size, max_len, num_classes)\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy',f1_score])\n",
    "\n",
    "batch_size = 256\n",
    "epochs = 10\n",
    "history = model.fit(X_train, y_train, batch_size = batch_size, epochs = 30, validation_split = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['acc', 'f1_score', 'loss', 'val_f1_score', 'val_acc', 'val_loss']\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6268b7c910>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626f01fe10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plotHistory(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test set accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.0314898933059855, 0.70013839768094865, 0.32392416223276399]\n"
     ]
    }
   ],
   "source": [
    "#loss, acc = model.evaluate(__X, __Y, verbose=0)\n",
    "print(model.evaluate(__X, __Y, verbose=0))\n",
    "\n",
    "#np.argmax(model.predict(__X),axis=1), _Y"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "__pred = model.predict(X_test).argmax(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    label  count\n",
      "0       0     44\n",
      "1       1    250\n",
      "2       2    146\n",
      "3       3    313\n",
      "4       4    192\n",
      "5       5    185\n",
      "6       6    338\n",
      "7       7    639\n",
      "8       8    326\n",
      "9       9    115\n",
      "10     10    367\n",
      "11     11     85\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f626f06b7d0>"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626ebdc2d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "df = pd.DataFrame(__pred, columns=[\"label\"]).groupby([\"label\"]).size().to_frame(name = 'count').reset_index()\n",
    "print df\n",
    "df.plot(kind=\"bar\", x = \"label\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Same the predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "f = open('pred.txt', 'w')\n",
    "for item in __pred:\n",
    "  f.write(\"%s\\n\" % item)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Original model without embedding layer (One hot encoding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train_encode = OneHotEncode().transform(X_train)\n",
    "x_test_encode = OneHotEncode().transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def originalModelWithoutEmb(input_shape, num_classes):\n",
    "    \n",
    "    kernel_size = 5\n",
    "    filters = 64\n",
    "    pool_size = 4\n",
    "    lstm_output_size = 200\n",
    "    \n",
    "    model = Sequential()\n",
    "    #model.add(Embedding(max_features, embedding_size, input_length=maxlen))\n",
    "    #model.add(BatchNormalization())\n",
    "    #model.add(Dropout(0.20))\n",
    "    #model.add(Conv1D(filters=filters, kernel_size = kernel_size, padding='valid', activation='relu', strides=1))\n",
    "    model.add(Conv1D(filters= filters, kernel_size= kernel_size, padding='valid', activation='relu', strides=1, input_shape=input_shape))\n",
    "    model.add(MaxPooling1D(pool_size=pool_size))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 5, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 6, padding='valid', activation='relu', strides=1))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Conv1D(filters=filters, kernel_size = 8, padding='valid', activation='relu', strides=1))\n",
    "    model.add(LSTM(lstm_output_size))\n",
    "    model.add(Dense(100))\n",
    "    model.add(Dense(50))\n",
    "    model.add(Dense(num_classes))\n",
    "    model.add(Activation('sigmoid'))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 23409 samples, validate on 2601 samples\n",
      "Epoch 1/10\n",
      "23409/23409 [==============================] - 42s 2ms/step - loss: 2.3535 - acc: 0.1582 - val_loss: 2.3456 - val_acc: 0.1499\n",
      "Epoch 2/10\n",
      "23409/23409 [==============================] - 42s 2ms/step - loss: 2.2555 - acc: 0.1828 - val_loss: 2.1574 - val_acc: 0.2103\n",
      "Epoch 3/10\n",
      "23409/23409 [==============================] - 39s 2ms/step - loss: 2.0688 - acc: 0.2314 - val_loss: 2.0158 - val_acc: 0.2280\n",
      "Epoch 4/10\n",
      "23409/23409 [==============================] - 41s 2ms/step - loss: 1.9423 - acc: 0.2707 - val_loss: 1.8965 - val_acc: 0.2730\n",
      "Epoch 5/10\n",
      "23409/23409 [==============================] - 40s 2ms/step - loss: 1.8724 - acc: 0.2845 - val_loss: 1.7777 - val_acc: 0.3337\n",
      "Epoch 6/10\n",
      "23409/23409 [==============================] - 36s 2ms/step - loss: 1.7020 - acc: 0.3627 - val_loss: 1.6120 - val_acc: 0.4229\n",
      "Epoch 7/10\n",
      "23409/23409 [==============================] - 36s 2ms/step - loss: 1.5369 - acc: 0.4371 - val_loss: 1.4957 - val_acc: 0.4648\n",
      "Epoch 8/10\n",
      "23409/23409 [==============================] - 36s 2ms/step - loss: 1.4268 - acc: 0.4752 - val_loss: 1.6034 - val_acc: 0.4371\n",
      "Epoch 9/10\n",
      "23409/23409 [==============================] - 37s 2ms/step - loss: 1.3542 - acc: 0.5084 - val_loss: 1.3730 - val_acc: 0.5148\n",
      "Epoch 10/10\n",
      "23409/23409 [==============================] - 36s 2ms/step - loss: 1.2918 - acc: 0.5315 - val_loss: 1.3379 - val_acc: 0.5225\n"
     ]
    }
   ],
   "source": [
    "model2 = originalModelWithoutEmb(x_train_encode[0].shape, num_classes)\n",
    "model2.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "batch_size = 256 \n",
    "epochs = 10\n",
    "history = model2.fit(x_train_encode, y_train, batch_size = batch_size, epochs = epochs, validation_split = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['acc', 'loss', 'val_acc', 'val_loss']\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fe6568a7ed0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fe6568e99d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plotHistory(history)"
   ]
  }
 ],
 "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": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}