lstm_autoencoder_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "CC5S3BbZKunc"
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from pylab import rcParams\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import optimizers, Sequential\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.utils import plot_model\n",
    "from tensorflow.keras.layers import Dense, LSTM, RepeatVector, TimeDistributed\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, precision_recall_curve\n",
    "from sklearn.metrics import recall_score, classification_report, auc, roc_curve\n",
    "from sklearn.metrics import precision_recall_fscore_support, f1_score\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(7)\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "P33wTOUlKw7W"
   },
   "outputs": [],
   "source": [
    "def export_csv():\n",
    "    # Left lower extremity\n",
    "    res = {}\n",
    "\n",
    "    # # Annotations and Metadata\n",
    "    # res['GRF_Annotation'] = pd.read_csv(r'https://s3-eu-west-1.amazonaws.com/pstorage-npg-968563215/22062960/GRF_metadata.csv')\n",
    "\n",
    "    # Left lower extremity\n",
    "    #res['GRF_F_V_PRO_left'] = pd.read_csv(r'GRF_F_V_PRO_left.csv')\n",
    "\n",
    "    #res['GRF_F_AP_PRO_left'] = pd.read_csv(r'GRF_F_AP_PRO_left.csv')\n",
    "\n",
    "    #res['GRF_F_ML_PRO_left'] = pd.read_csv(r'GRF_F_ML_PRO_left.csv')\n",
    "\n",
    "    # res['GRF_COP_AP_PRO_left'] = pd.read_csv(r'GRF_COP_AP_PRO_left.csv')\n",
    "\n",
    "    # res['GRF_COP_ML_PRO_left'] = pd.read_csv(r'GRF_COP_ML_PRO_left.csv')\n",
    "\n",
    "    # Right lower extremity\n",
    "    #res['GRF_F_V_PRO_right'] = pd.read_csv(r'GRF_F_V_PRO_right.csv')\n",
    "\n",
    "    #res['GRF_F_AP_PRO_right'] = pd.read_csv(r'GRF_F_AP_PRO_right.csv')\n",
    "\n",
    "    res['GRF_F_ML_PRO_right'] = pd.read_csv(r'GRF_F_ML_PRO_right.csv')\n",
    "\n",
    "    #res['GRF_COP_AP_PRO_right'] = pd.read_csv(r'GRF_COP_AP_PRO_right.csv')\n",
    "\n",
    "    #res['GRF_COP_ML_PRO_right'] = pd.read_csv(r'GRF_COP_ML_PRO_right.csv')\n",
    "\n",
    "    # Annotations and Metadata\n",
    "    res['GRF_Annotation'] = pd.read_csv(r'GRF_metadata.csv')\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "mYBwqnXkKzHz"
   },
   "outputs": [],
   "source": [
    "df_dict = export_csv()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "8wZIZOIeK1Yq"
   },
   "outputs": [],
   "source": [
    "def getTitle(label):\n",
    "    if label==\"HC\":\n",
    "        return \"Healthy Controls\"\n",
    "    elif label==\"K\":\n",
    "        return \"Knee impairment \"\n",
    "    elif label==\"A\":\n",
    "        return \"Ankle impairment\"\n",
    "    elif label==\"H\":\n",
    "        return \"Hip impairment\"\n",
    "    else:\n",
    "        return \"Calcaneus impairment\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "ltWiMIj7LCzc"
   },
   "outputs": [],
   "source": [
    "def get_Inputs(grf, meta, label):\n",
    "    subject_id = meta.loc[meta[\"CLASS_LABEL\"]==label][\"SUBJECT_ID\"].unique()\n",
    "    subject_id = pd.DataFrame({\"SUBJECT_ID\": subject_id})\n",
    "    grf = subject_id.merge(grf, on=\"SUBJECT_ID\", how = \"inner\")\n",
    "    #grf = grf.drop_duplicates(['SUBJECT_ID'], keep = 'first')\n",
    "    #grf.drop(grf.columns[:3], axis = 1)\n",
    "    grf = grf.iloc[:,3:].values\n",
    "    res = np.array(grf)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "S4DqcgVuLEnL",
    "outputId": "f68adb1d-d97d-4e9a-f2a2-379f8ac7e45a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7755, 101)\n"
     ]
    }
   ],
   "source": [
    "X = None\n",
    "for key in df_dict:\n",
    "    if key== 'GRF_Annotation': \n",
    "        continue\n",
    "    temp = get_Inputs(df_dict[key], df_dict['GRF_Annotation'], 'HC') #data of healthy controls is used for training\n",
    "    if X is None:\n",
    "        X = temp\n",
    "    else:\n",
    "        X = np.append(X, temp, axis=0)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "xyykcP_JZthn",
    "outputId": "b7b009e1-24da-495d-944c-c483a75e0b1a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4653, 101)\n",
      "(2326, 101)\n",
      "(776, 101)\n"
     ]
    }
   ],
   "source": [
    "#scaling and reshaping of the input data for the model\n",
    "scaler = MinMaxScaler(feature_range=(0, 1))\n",
    "#print('X', X)\n",
    "X = scaler.fit_transform(X)\n",
    "#print('X_new', X)\n",
    "l1, l2 = 0.6, 0.9\n",
    "X_train = X[0:(int)(X.shape[0]*l1), :]\n",
    "X_val = X[(int)(X.shape[0]*l1):(int)(X.shape[0]*l2), :]\n",
    "X_test = X[(int)(X.shape[0]*l2):, :]\n",
    "\n",
    "print(X_train.shape)\n",
    "print(X_val.shape)\n",
    "print(X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "LHOxHi5xLHnt"
   },
   "outputs": [],
   "source": [
    "#X = get_Inputs(df_dict['GRF_F_V_PRO_left'], df_dict['GRF_Annotation'], 'A')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "usXI576ILJgD"
   },
   "outputs": [],
   "source": [
    "def get_data_series(X, serie_size):\n",
    "    #first_step = 0 #X.shape[1] - serie_size - 1\n",
    "    print(X.shape)\n",
    "    last_step = X.shape[1]\n",
    "    data_series = []\n",
    "    for row in X:\n",
    "        for step1 in range(last_step - serie_size):\n",
    "            serie = row[step1 : step1 + serie_size]\n",
    "            # for step2 in range(serie_size):\n",
    "            #     serie.append(row[step1 + step2])\n",
    "            data_series.append(serie)\n",
    "    res = np.array(np.squeeze(np.array(data_series)))\n",
    "    print(res.shape)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "jsoLO-eHLK_j",
    "outputId": "4d23195a-71f0-42d4-ccb0-504f81b324c2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4653, 101)\n",
      "(395505, 16)\n",
      "(2326, 101)\n",
      "(197710, 16)\n",
      "(776, 101)\n",
      "(65960, 16)\n",
      "X_train.shape  (395505, 16, 1)\n",
      "X_val.shape  (197710, 16, 1)\n",
      "X_test.shape  (65960, 16, 1)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_val, X_test = get_data_series(X_train, 16),get_data_series(X_val, 16) ,get_data_series(X_test, 16)\n",
    "\n",
    "X_train, X_val, X_test = X_train.reshape(X_train.shape[0], X_train.shape[1], 1),X_val.reshape(X_val.shape[0], X_val.shape[1], 1), X_test.reshape(X_test.shape[0], X_test.shape[1], 1)\n",
    "print('X_train.shape ',X_train.shape)\n",
    "print('X_val.shape ',X_val.shape)\n",
    "print('X_test.shape ',X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "akaYzUllLO0J"
   },
   "outputs": [],
   "source": [
    "serie_size  =  X_train.shape[1] # equal to the lookback\n",
    "n_features =  X_train.shape[2] # 1\n",
    "\n",
    "epochs = 150\n",
    "batch = 128\n",
    "lr = 0.001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "FZSeYE9VVBqL",
    "outputId": "3517f9dc-e453-4708-a440-d2c0aa56f63e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "WARNING:tensorflow:Layer lstm_1 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "WARNING:tensorflow:Layer lstm_2 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "WARNING:tensorflow:Layer lstm_3 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "WARNING:tensorflow:Layer lstm_4 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, 16, 16)            1152      \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 16, 8)             800       \n",
      "_________________________________________________________________\n",
      "lstm_2 (LSTM)                (None, 4)                 208       \n",
      "_________________________________________________________________\n",
      "repeat_vector (RepeatVector) (None, 16, 4)             0         \n",
      "_________________________________________________________________\n",
      "lstm_3 (LSTM)                (None, 16, 8)             416       \n",
      "_________________________________________________________________\n",
      "lstm_4 (LSTM)                (None, 16, 16)            1600      \n",
      "_________________________________________________________________\n",
      "time_distributed (TimeDistri (None, 16, 1)             17        \n",
      "=================================================================\n",
      "Total params: 4,193\n",
      "Trainable params: 4,193\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#using LSTM-AE\n",
    "lstm_autoencoder = Sequential()\n",
    "# Encoder\n",
    "lstm_autoencoder.add(LSTM(serie_size, activation='relu', input_shape=(serie_size, n_features), return_sequences=True))\n",
    "lstm_autoencoder.add(LSTM(8, activation='relu', return_sequences=True))\n",
    "#lstm_autoencoder.add(LSTM(8, activation='relu'))\n",
    "lstm_autoencoder.add(LSTM(4, activation='relu'))\n",
    "lstm_autoencoder.add(RepeatVector(serie_size))\n",
    "# lstm_autoencoder.add(LSTM(32, activation='relu', return_sequences=True))\n",
    "# lstm_autoencoder.add(LSTM(16, activation='relu', return_sequences=False))\n",
    "# lstm_autoencoder.add(RepeatVector(serie_size))\n",
    "# # Decoder\n",
    "# lstm_autoencoder.add(LSTM(16, activation='relu', return_sequences=True))\n",
    "# lstm_autoencoder.add(LSTM(32, activation='relu', return_sequences=True))\n",
    "lstm_autoencoder.add(LSTM(8, activation='relu', return_sequences=True))\n",
    "#lstm_autoencoder.add(LSTM(16, activation='relu', return_sequences=True))\n",
    "lstm_autoencoder.add(LSTM(serie_size, activation='relu', return_sequences=True))\n",
    "lstm_autoencoder.add(TimeDistributed(Dense(n_features)))\n",
    "\n",
    "lstm_autoencoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/150\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x0000019F00F25F28> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 4, expecting 3\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x0000019F00F25F28> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 4, expecting 3\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x0000019F48EB3840> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 4, expecting 3\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x0000019F48EB3840> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 4, expecting 3\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "3090/3090 - 476s - loss: 0.0093 - val_loss: 0.0012\n",
      "Epoch 2/150\n",
      "3090/3090 - 557s - loss: 9.6107e-04 - val_loss: 6.0701e-04\n",
      "Epoch 3/150\n",
      "3090/3090 - 620s - loss: 3.9363e-04 - val_loss: 3.3462e-04\n",
      "Epoch 4/150\n",
      "3090/3090 - 612s - loss: 3.5571e-04 - val_loss: 3.3823e-04\n",
      "Epoch 5/150\n",
      "3090/3090 - 603s - loss: 3.4373e-04 - val_loss: 3.3592e-04\n",
      "Epoch 6/150\n",
      "3090/3090 - 606s - loss: 3.3518e-04 - val_loss: 3.4415e-04\n",
      "Epoch 7/150\n",
      "3090/3090 - 612s - loss: 3.3046e-04 - val_loss: 3.4741e-04\n",
      "Epoch 8/150\n",
      "3090/3090 - 606s - loss: 3.2666e-04 - val_loss: 3.2989e-04\n",
      "Epoch 9/150\n",
      "3090/3090 - 575s - loss: 3.2339e-04 - val_loss: 3.2382e-04\n",
      "Epoch 10/150\n",
      "3090/3090 - 579s - loss: 3.2006e-04 - val_loss: 3.0221e-04\n",
      "Epoch 11/150\n",
      "3090/3090 - 565s - loss: 3.1783e-04 - val_loss: 3.0295e-04\n",
      "Epoch 12/150\n",
      "3090/3090 - 561s - loss: 3.1493e-04 - val_loss: 2.9295e-04\n",
      "Epoch 13/150\n",
      "3090/3090 - 563s - loss: 3.1255e-04 - val_loss: 3.1628e-04\n",
      "Epoch 14/150\n",
      "3090/3090 - 672s - loss: 3.0766e-04 - val_loss: 2.8105e-04\n",
      "Epoch 15/150\n",
      "3090/3090 - 729s - loss: 2.9806e-04 - val_loss: 2.5407e-04\n",
      "Epoch 16/150\n",
      "3090/3090 - 763s - loss: 2.7246e-04 - val_loss: 2.2801e-04\n",
      "Epoch 17/150\n",
      "3090/3090 - 589s - loss: 2.3379e-04 - val_loss: 2.0088e-04\n",
      "Epoch 18/150\n",
      "3090/3090 - 578s - loss: 2.0291e-04 - val_loss: 2.0643e-04\n",
      "Epoch 19/150\n",
      "3090/3090 - 577s - loss: 1.8039e-04 - val_loss: 1.4788e-04\n",
      "Epoch 20/150\n",
      "3090/3090 - 579s - loss: 1.6828e-04 - val_loss: 1.7870e-04\n",
      "Epoch 21/150\n",
      "3090/3090 - 574s - loss: 1.6052e-04 - val_loss: 1.5568e-04\n",
      "Epoch 22/150\n",
      "3090/3090 - 576s - loss: 1.5538e-04 - val_loss: 1.6946e-04\n",
      "Epoch 23/150\n",
      "3090/3090 - 577s - loss: 1.5146e-04 - val_loss: 1.2878e-04\n",
      "Epoch 24/150\n",
      "3090/3090 - 575s - loss: 1.4824e-04 - val_loss: 1.8398e-04\n",
      "Epoch 25/150\n",
      "3090/3090 - 607s - loss: 1.4589e-04 - val_loss: 1.3732e-04\n",
      "Epoch 26/150\n",
      "3090/3090 - 593s - loss: 1.4342e-04 - val_loss: 1.2932e-04\n",
      "Epoch 27/150\n",
      "3090/3090 - 591s - loss: 1.4120e-04 - val_loss: 1.2719e-04\n",
      "Epoch 28/150\n",
      "3090/3090 - 618s - loss: 1.3947e-04 - val_loss: 1.2841e-04\n",
      "Epoch 29/150\n",
      "3090/3090 - 582s - loss: 1.3806e-04 - val_loss: 1.2188e-04\n",
      "Epoch 30/150\n",
      "3090/3090 - 576s - loss: 1.3648e-04 - val_loss: 1.2490e-04\n",
      "Epoch 31/150\n",
      "3090/3090 - 601s - loss: 1.3553e-04 - val_loss: 1.3481e-04\n",
      "Epoch 32/150\n",
      "3090/3090 - 614s - loss: 1.3414e-04 - val_loss: 1.3315e-04\n",
      "Epoch 33/150\n",
      "3090/3090 - 638s - loss: 1.3326e-04 - val_loss: 1.4339e-04\n",
      "Epoch 34/150\n",
      "3090/3090 - 592s - loss: 1.3249e-04 - val_loss: 1.4144e-04\n",
      "Epoch 35/150\n",
      "3090/3090 - 577s - loss: 1.3158e-04 - val_loss: 1.2429e-04\n",
      "Epoch 36/150\n",
      "3090/3090 - 614s - loss: 1.3101e-04 - val_loss: 1.1939e-04\n",
      "Epoch 37/150\n",
      "3090/3090 - 571s - loss: 1.3003e-04 - val_loss: 1.2347e-04\n",
      "Epoch 38/150\n",
      "3090/3090 - 577s - loss: 1.2969e-04 - val_loss: 1.3599e-04\n",
      "Epoch 39/150\n",
      "3090/3090 - 567s - loss: 1.2860e-04 - val_loss: 1.1897e-04\n",
      "Epoch 40/150\n",
      "3090/3090 - 565s - loss: 1.2829e-04 - val_loss: 1.4158e-04\n",
      "Epoch 41/150\n",
      "3090/3090 - 563s - loss: 1.2762e-04 - val_loss: 1.1609e-04\n",
      "Epoch 42/150\n",
      "3090/3090 - 572s - loss: 1.2682e-04 - val_loss: 1.1803e-04\n",
      "Epoch 43/150\n",
      "3090/3090 - 562s - loss: 1.2640e-04 - val_loss: 1.1758e-04\n",
      "Epoch 44/150\n",
      "3090/3090 - 562s - loss: 1.2589e-04 - val_loss: 1.2911e-04\n",
      "Epoch 45/150\n",
      "3090/3090 - 562s - loss: 1.2546e-04 - val_loss: 1.1712e-04\n",
      "Epoch 46/150\n",
      "3090/3090 - 562s - loss: 1.2508e-04 - val_loss: 1.3675e-04\n",
      "Epoch 47/150\n",
      "3090/3090 - 563s - loss: 1.2459e-04 - val_loss: 1.2375e-04\n",
      "Epoch 48/150\n",
      "3090/3090 - 567s - loss: 1.2415e-04 - val_loss: 1.1694e-04\n",
      "Epoch 49/150\n",
      "3090/3090 - 562s - loss: 1.2375e-04 - val_loss: 1.1853e-04\n",
      "Epoch 50/150\n",
      "3090/3090 - 563s - loss: 1.2343e-04 - val_loss: 1.1378e-04\n",
      "Epoch 51/150\n",
      "3090/3090 - 563s - loss: 1.2305e-04 - val_loss: 1.1989e-04\n",
      "Epoch 52/150\n",
      "3090/3090 - 567s - loss: 1.2277e-04 - val_loss: 1.1420e-04\n",
      "Epoch 53/150\n",
      "3090/3090 - 572s - loss: 1.2251e-04 - val_loss: 1.2073e-04\n",
      "Epoch 54/150\n",
      "3090/3090 - 569s - loss: 1.2204e-04 - val_loss: 2.0649e-04\n",
      "Epoch 55/150\n",
      "3090/3090 - 567s - loss: 1.2189e-04 - val_loss: 1.3610e-04\n",
      "Epoch 56/150\n",
      "3090/3090 - 567s - loss: 1.2155e-04 - val_loss: 1.1627e-04\n",
      "Epoch 57/150\n",
      "3090/3090 - 567s - loss: 1.2132e-04 - val_loss: 1.1877e-04\n",
      "Epoch 58/150\n",
      "3090/3090 - 590s - loss: 1.2110e-04 - val_loss: 1.1720e-04\n",
      "Epoch 59/150\n",
      "3090/3090 - 567s - loss: 1.2085e-04 - val_loss: 1.1839e-04\n",
      "Epoch 60/150\n",
      "3090/3090 - 571s - loss: 1.2047e-04 - val_loss: 1.2112e-04\n",
      "Epoch 61/150\n",
      "3090/3090 - 564s - loss: 1.2038e-04 - val_loss: 1.1587e-04\n",
      "Epoch 62/150\n",
      "3090/3090 - 569s - loss: 1.2014e-04 - val_loss: 1.3168e-04\n",
      "Epoch 63/150\n",
      "3090/3090 - 562s - loss: 1.2000e-04 - val_loss: 1.1813e-04\n",
      "Epoch 64/150\n",
      "3090/3090 - 562s - loss: 1.1968e-04 - val_loss: 1.2631e-04\n",
      "Epoch 65/150\n",
      "3090/3090 - 562s - loss: 1.1953e-04 - val_loss: 1.2125e-04\n",
      "Epoch 66/150\n",
      "3090/3090 - 567s - loss: 1.1927e-04 - val_loss: 1.2550e-04\n",
      "Epoch 67/150\n",
      "3090/3090 - 564s - loss: 1.1915e-04 - val_loss: 1.1470e-04\n",
      "Epoch 68/150\n",
      "3090/3090 - 566s - loss: 1.1895e-04 - val_loss: 1.1432e-04\n",
      "Epoch 69/150\n",
      "3090/3090 - 566s - loss: 1.1868e-04 - val_loss: 1.2767e-04\n",
      "Epoch 70/150\n",
      "3090/3090 - 566s - loss: 1.1859e-04 - val_loss: 1.1218e-04\n",
      "Epoch 71/150\n",
      "3090/3090 - 565s - loss: 1.1842e-04 - val_loss: 1.1159e-04\n",
      "Epoch 72/150\n",
      "3090/3090 - 564s - loss: 1.1827e-04 - val_loss: 1.1341e-04\n",
      "Epoch 73/150\n",
      "3090/3090 - 566s - loss: 1.1793e-04 - val_loss: 1.1354e-04\n",
      "Epoch 74/150\n",
      "3090/3090 - 566s - loss: 1.1779e-04 - val_loss: 1.4139e-04\n",
      "Epoch 75/150\n",
      "3090/3090 - 557s - loss: 1.1772e-04 - val_loss: 1.1225e-04\n",
      "Epoch 76/150\n",
      "3090/3090 - 566s - loss: 1.1757e-04 - val_loss: 1.1519e-04\n",
      "Epoch 77/150\n",
      "3090/3090 - 568s - loss: 1.1738e-04 - val_loss: 1.1337e-04\n",
      "Epoch 78/150\n",
      "3090/3090 - 564s - loss: 1.1720e-04 - val_loss: 1.1733e-04\n",
      "Epoch 79/150\n",
      "3090/3090 - 571s - loss: 1.1715e-04 - val_loss: 1.1186e-04\n",
      "Epoch 80/150\n",
      "3090/3090 - 565s - loss: 1.1688e-04 - val_loss: 1.1278e-04\n",
      "Epoch 81/150\n",
      "3090/3090 - 565s - loss: 1.1679e-04 - val_loss: 1.2906e-04\n",
      "Epoch 82/150\n",
      "3090/3090 - 559s - loss: 1.1675e-04 - val_loss: 1.1181e-04\n",
      "Epoch 83/150\n",
      "3090/3090 - 568s - loss: 1.1651e-04 - val_loss: 1.1074e-04\n",
      "Epoch 84/150\n",
      "3090/3090 - 563s - loss: 1.1640e-04 - val_loss: 1.1312e-04\n",
      "Epoch 85/150\n",
      "3090/3090 - 560s - loss: 1.1620e-04 - val_loss: 1.1440e-04\n",
      "Epoch 86/150\n",
      "3090/3090 - 565s - loss: 1.1605e-04 - val_loss: 1.1965e-04\n",
      "Epoch 87/150\n",
      "3090/3090 - 562s - loss: 1.1596e-04 - val_loss: 1.1186e-04\n",
      "Epoch 88/150\n",
      "3090/3090 - 562s - loss: 1.1593e-04 - val_loss: 1.1472e-04\n",
      "Epoch 89/150\n",
      "3090/3090 - 563s - loss: 1.1579e-04 - val_loss: 1.1217e-04\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 90/150\n",
      "3090/3090 - 560s - loss: 1.1567e-04 - val_loss: 1.1682e-04\n",
      "Epoch 91/150\n",
      "3090/3090 - 560s - loss: 1.1542e-04 - val_loss: 1.1738e-04\n",
      "Epoch 92/150\n",
      "3090/3090 - 567s - loss: 1.1546e-04 - val_loss: 1.1118e-04\n",
      "Epoch 93/150\n",
      "3090/3090 - 573s - loss: 1.1533e-04 - val_loss: 1.1933e-04\n",
      "Epoch 94/150\n",
      "3090/3090 - 564s - loss: 1.1522e-04 - val_loss: 1.1774e-04\n",
      "Epoch 95/150\n",
      "3090/3090 - 580s - loss: 1.1508e-04 - val_loss: 1.1105e-04\n",
      "Epoch 96/150\n",
      "3090/3090 - 633s - loss: 1.1482e-04 - val_loss: 1.1553e-04\n",
      "Epoch 97/150\n",
      "3090/3090 - 658s - loss: 1.1477e-04 - val_loss: 1.1071e-04\n",
      "Epoch 98/150\n",
      "3090/3090 - 660s - loss: 1.1466e-04 - val_loss: 1.1228e-04\n",
      "Epoch 99/150\n",
      "3090/3090 - 603s - loss: 1.1449e-04 - val_loss: 1.1199e-04\n",
      "Epoch 100/150\n",
      "3090/3090 - 621s - loss: 1.1446e-04 - val_loss: 1.0963e-04\n",
      "Epoch 101/150\n",
      "3090/3090 - 681s - loss: 1.1427e-04 - val_loss: 1.1266e-04\n",
      "Epoch 102/150\n",
      "3090/3090 - 619s - loss: 1.1404e-04 - val_loss: 1.1209e-04\n",
      "Epoch 103/150\n",
      "3090/3090 - 648s - loss: 1.1409e-04 - val_loss: 1.1431e-04\n",
      "Epoch 104/150\n",
      "3090/3090 - 688s - loss: 1.1390e-04 - val_loss: 1.2094e-04\n",
      "Epoch 105/150\n",
      "3090/3090 - 652s - loss: 1.1374e-04 - val_loss: 1.1265e-04\n",
      "Epoch 106/150\n",
      "3090/3090 - 642s - loss: 1.1360e-04 - val_loss: 1.1417e-04\n",
      "Epoch 107/150\n",
      "3090/3090 - 685s - loss: 1.1358e-04 - val_loss: 1.1062e-04\n",
      "Epoch 108/150\n",
      "3090/3090 - 658s - loss: 1.1347e-04 - val_loss: 1.1112e-04\n",
      "Epoch 109/150\n",
      "3090/3090 - 691s - loss: 1.1340e-04 - val_loss: 1.0920e-04\n",
      "Epoch 110/150\n",
      "3090/3090 - 674s - loss: 1.1322e-04 - val_loss: 1.1306e-04\n",
      "Epoch 111/150\n",
      "3090/3090 - 621s - loss: 1.1301e-04 - val_loss: 1.1095e-04\n",
      "Epoch 112/150\n",
      "3090/3090 - 661s - loss: 1.1298e-04 - val_loss: 1.1132e-04\n",
      "Epoch 113/150\n",
      "3090/3090 - 677s - loss: 1.1289e-04 - val_loss: 1.0914e-04\n",
      "Epoch 114/150\n",
      "3090/3090 - 673s - loss: 1.1272e-04 - val_loss: 1.1219e-04\n",
      "Epoch 115/150\n",
      "3090/3090 - 680s - loss: 1.1256e-04 - val_loss: 1.1061e-04\n",
      "Epoch 116/150\n",
      "3090/3090 - 679s - loss: 1.1255e-04 - val_loss: 1.1058e-04\n",
      "Epoch 117/150\n",
      "3090/3090 - 677s - loss: 1.1241e-04 - val_loss: 1.1252e-04\n",
      "Epoch 118/150\n",
      "3090/3090 - 669s - loss: 1.1231e-04 - val_loss: 1.0980e-04\n",
      "Epoch 119/150\n",
      "3090/3090 - 675s - loss: 1.1219e-04 - val_loss: 1.0865e-04\n",
      "Epoch 120/150\n",
      "3090/3090 - 643s - loss: 1.1213e-04 - val_loss: 1.0875e-04\n",
      "Epoch 121/150\n",
      "3090/3090 - 669s - loss: 1.1206e-04 - val_loss: 1.1087e-04\n",
      "Epoch 122/150\n",
      "3090/3090 - 678s - loss: 1.1185e-04 - val_loss: 1.0907e-04\n",
      "Epoch 123/150\n",
      "3090/3090 - 626s - loss: 1.1186e-04 - val_loss: 1.1073e-04\n",
      "Epoch 124/150\n",
      "3090/3090 - 681s - loss: 1.1177e-04 - val_loss: 1.1251e-04\n",
      "Epoch 125/150\n",
      "3090/3090 - 671s - loss: 1.1172e-04 - val_loss: 1.0841e-04\n",
      "Epoch 126/150\n",
      "3090/3090 - 620s - loss: 1.1165e-04 - val_loss: 1.1186e-04\n",
      "Epoch 127/150\n",
      "3090/3090 - 686s - loss: 1.1147e-04 - val_loss: 1.0830e-04\n",
      "Epoch 128/150\n",
      "3090/3090 - 670s - loss: 1.1145e-04 - val_loss: 1.0899e-04\n",
      "Epoch 129/150\n",
      "3090/3090 - 660s - loss: 1.1132e-04 - val_loss: 1.1120e-04\n",
      "Epoch 130/150\n",
      "3090/3090 - 681s - loss: 1.1130e-04 - val_loss: 1.1046e-04\n",
      "Epoch 131/150\n",
      "3090/3090 - 635s - loss: 1.1129e-04 - val_loss: 1.0777e-04\n",
      "Epoch 132/150\n",
      "3090/3090 - 682s - loss: 1.1110e-04 - val_loss: 1.1780e-04\n",
      "Epoch 133/150\n",
      "3090/3090 - 658s - loss: 1.1104e-04 - val_loss: 1.0936e-04\n",
      "Epoch 134/150\n",
      "3090/3090 - 628s - loss: 1.1103e-04 - val_loss: 1.1192e-04\n",
      "Epoch 135/150\n",
      "3090/3090 - 627s - loss: 1.1090e-04 - val_loss: 1.0752e-04\n",
      "Epoch 136/150\n",
      "3090/3090 - 631s - loss: 1.1084e-04 - val_loss: 1.0907e-04\n",
      "Epoch 137/150\n",
      "3090/3090 - 617s - loss: 1.1070e-04 - val_loss: 1.1072e-04\n",
      "Epoch 138/150\n",
      "3090/3090 - 581s - loss: 1.1068e-04 - val_loss: 1.0860e-04\n",
      "Epoch 139/150\n",
      "3090/3090 - 574s - loss: 1.1060e-04 - val_loss: 1.0856e-04\n",
      "Epoch 140/150\n",
      "3090/3090 - 569s - loss: 1.1057e-04 - val_loss: 1.1096e-04\n",
      "Epoch 141/150\n",
      "3090/3090 - 567s - loss: 1.1051e-04 - val_loss: 1.0803e-04\n",
      "Epoch 142/150\n",
      "3090/3090 - 569s - loss: 1.1039e-04 - val_loss: 1.0868e-04\n",
      "Epoch 143/150\n",
      "3090/3090 - 567s - loss: 1.1045e-04 - val_loss: 1.0836e-04\n",
      "Epoch 144/150\n",
      "3090/3090 - 569s - loss: 1.1022e-04 - val_loss: 1.1187e-04\n",
      "Epoch 145/150\n",
      "3090/3090 - 615s - loss: 1.1025e-04 - val_loss: 1.0823e-04\n",
      "Epoch 146/150\n",
      "3090/3090 - 591s - loss: 1.1019e-04 - val_loss: 1.0854e-04\n",
      "Epoch 147/150\n",
      "3090/3090 - 580s - loss: 1.0998e-04 - val_loss: 1.0800e-04\n",
      "Epoch 148/150\n",
      "3090/3090 - 592s - loss: 1.1006e-04 - val_loss: 1.1023e-04\n",
      "Epoch 149/150\n",
      "3090/3090 - 581s - loss: 1.0992e-04 - val_loss: 1.0930e-04\n",
      "Epoch 150/150\n",
      "3090/3090 - 602s - loss: 1.0996e-04 - val_loss: 1.0805e-04\n"
     ]
    }
   ],
   "source": [
    "adam = optimizers.Adam(lr)\n",
    "lstm_autoencoder.compile(loss='mse', optimizer=adam)\n",
    "\n",
    "tb = TensorBoard(log_dir='./logs',\n",
    "                histogram_freq=0,\n",
    "                write_graph=True,\n",
    "                write_images=True)\n",
    "\n",
    "lstm_autoencoder_history = lstm_autoencoder.fit(X_train, X_train, \n",
    "                                                epochs=epochs, \n",
    "                                                batch_size=batch,\n",
    "                                                validation_data=(X_val, X_val),\n",
    "                                                verbose=2).history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib\n",
    "\n",
    "SMALL_SIZE = 20\n",
    "MEDIUM_SIZE = 25\n",
    "BIGGER_SIZE = 30\n",
    "#matplotlib.rc('font', size=SMALL_SIZE)\n",
    "matplotlib.rc('axes', titlesize=BIGGER_SIZE)\n",
    "matplotlib.rc('axes', labelsize=BIGGER_SIZE)\n",
    "matplotlib.rc('xtick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "matplotlib.rc('ytick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "matplotlib.rc('legend', fontsize=SMALL_SIZE)    # legend fontsize\n",
    "matplotlib.rc('figure', titlesize=BIGGER_SIZE) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "start = 1\n",
    "plt.figure(figsize= (10,10))\n",
    "plt.ticklabel_format(style='sci', axis='y', scilimits=(0,0))\n",
    "loss_train = lstm_autoencoder_history['loss'][start:]\n",
    "plt.plot(np.arange(start, len(lstm_autoencoder_history['loss'])), loss_train, linewidth=2, label='Train')\n",
    "loss_valid = lstm_autoencoder_history['val_loss'][start:]\n",
    "plt.plot(np.arange(start, len(lstm_autoencoder_history['loss'])), loss_valid, linewidth=2, label='Valid')\n",
    "plt.legend(loc='upper right')\n",
    "#plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "#plt.savefig(\"loss_plot_right_ml_ae.png\")\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "plot_7c_lstm_autoencoder_all_data.ipynb",
   "provenance": []
  },
  "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}