data_preparing_MIT.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "2a4c4d58",
   "metadata": {
    "id": "2a4c4d58"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from keras.models import load_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "1fc2bee0",
   "metadata": {},
   "outputs": [],
   "source": [
    "input_step=10 #to change after to 25,50\n",
    "output_step=10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "1b644822",
   "metadata": {
    "id": "1b644822"
   },
   "outputs": [],
   "source": [
    "X1=np.load(\"X1_\"+str(input_step)+\".npy\")\n",
    "X2=np.load(\"X2_\"+str(input_step)+\".npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "0655e9db",
   "metadata": {
    "id": "0655e9db",
    "outputId": "749486aa-abcc-4d6c-90c1-9402b5a7ff0a",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(56936, 10, 6, 500)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X1.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "3402d472",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "3402d472",
    "outputId": "28005772-178a-4d01-8f0b-e9c16ca75ea7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(55002, 10, 6, 500)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X2.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "f5b241e9",
   "metadata": {
    "id": "f5b241e9"
   },
   "outputs": [],
   "source": [
    "X1=np.float32(X1)\n",
    "X2=np.float32(X2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "759cfbf6",
   "metadata": {
    "id": "759cfbf6",
    "outputId": "3dcc018a-9b59-4fac-a51d-54f5645a112b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "KerasTensor(type_spec=TensorSpec(shape=(None, 10, 6, 500), dtype=tf.float32, name='input_4'), name='input_4', description=\"created by layer 'input_4'\")\n",
      "Model: \"model_4\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_4 (InputLayer)        [(None, 10, 6, 500)]      0         \n",
      "                                                                 \n",
      " flatten_1 (Flatten)         (None, 30000)             0         \n",
      "                                                                 \n",
      " dense_9 (Dense)             (None, 256)               7680256   \n",
      "                                                                 \n",
      " dense_10 (Dense)            (None, 128)               32896     \n",
      "                                                                 \n",
      " dense_11 (Dense)            (None, 100)               12900     \n",
      "                                                                 \n",
      " dense_12 (Dense)            (None, 640)               64640     \n",
      "                                                                 \n",
      " dense_13 (Dense)            (None, 256)               164096    \n",
      "                                                                 \n",
      " dense_14 (Dense)            (None, 128)               32896     \n",
      "                                                                 \n",
      " dense_15 (Dense)            (None, 256)               33024     \n",
      "                                                                 \n",
      " dense_16 (Dense)            (None, 30000)             7710000   \n",
      "                                                                 \n",
      " reshape_5 (Reshape)         (None, 10, 6, 500)        0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 15,730,708\n",
      "Trainable params: 15,730,708\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/300\n",
      "890/890 [==============================] - 52s 58ms/step - loss: 0.0447 - mae: 0.0447 - mse: 0.0094 - root_mean_squared_error: 0.0970 - val_loss: 0.1051 - val_mae: 0.1051 - val_mse: 0.0430 - val_root_mean_squared_error: 0.2075\n",
      "Epoch 2/300\n",
      "890/890 [==============================] - 49s 55ms/step - loss: 0.0289 - mae: 0.0289 - mse: 0.0058 - root_mean_squared_error: 0.0763 - val_loss: 0.1026 - val_mae: 0.1026 - val_mse: 0.0427 - val_root_mean_squared_error: 0.2067\n",
      "Epoch 3/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0235 - mae: 0.0235 - mse: 0.0042 - root_mean_squared_error: 0.0647 - val_loss: 0.0943 - val_mae: 0.0943 - val_mse: 0.0370 - val_root_mean_squared_error: 0.1923\n",
      "Epoch 4/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0218 - mae: 0.0218 - mse: 0.0038 - root_mean_squared_error: 0.0613 - val_loss: 0.0857 - val_mae: 0.0857 - val_mse: 0.0325 - val_root_mean_squared_error: 0.1804\n",
      "Epoch 5/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0195 - mae: 0.0195 - mse: 0.0032 - root_mean_squared_error: 0.0565 - val_loss: 0.0806 - val_mae: 0.0806 - val_mse: 0.0288 - val_root_mean_squared_error: 0.1697\n",
      "Epoch 6/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0184 - mae: 0.0184 - mse: 0.0029 - root_mean_squared_error: 0.0540 - val_loss: 0.0796 - val_mae: 0.0796 - val_mse: 0.0286 - val_root_mean_squared_error: 0.1692\n",
      "Epoch 7/300\n",
      "890/890 [==============================] - 49s 55ms/step - loss: 0.0176 - mae: 0.0176 - mse: 0.0028 - root_mean_squared_error: 0.0526 - val_loss: 0.0794 - val_mae: 0.0794 - val_mse: 0.0291 - val_root_mean_squared_error: 0.1706\n",
      "Epoch 8/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0167 - mae: 0.0167 - mse: 0.0026 - root_mean_squared_error: 0.0511 - val_loss: 0.0774 - val_mae: 0.0774 - val_mse: 0.0286 - val_root_mean_squared_error: 0.1692\n",
      "Epoch 9/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0161 - mae: 0.0161 - mse: 0.0025 - root_mean_squared_error: 0.0501 - val_loss: 0.0764 - val_mae: 0.0764 - val_mse: 0.0281 - val_root_mean_squared_error: 0.1676\n",
      "Epoch 10/300\n",
      "890/890 [==============================] - 49s 55ms/step - loss: 0.0157 - mae: 0.0157 - mse: 0.0024 - root_mean_squared_error: 0.0495 - val_loss: 0.0753 - val_mae: 0.0753 - val_mse: 0.0274 - val_root_mean_squared_error: 0.1655\n",
      "Epoch 11/300\n",
      "890/890 [==============================] - 49s 56ms/step - loss: 0.0154 - mae: 0.0154 - mse: 0.0024 - root_mean_squared_error: 0.0489 - val_loss: 0.0722 - val_mae: 0.0722 - val_mse: 0.0256 - val_root_mean_squared_error: 0.1598\n",
      "Epoch 12/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0151 - mae: 0.0151 - mse: 0.0023 - root_mean_squared_error: 0.0483 - val_loss: 0.0718 - val_mae: 0.0718 - val_mse: 0.0252 - val_root_mean_squared_error: 0.1588\n",
      "Epoch 13/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0147 - mae: 0.0147 - mse: 0.0023 - root_mean_squared_error: 0.0478 - val_loss: 0.0689 - val_mae: 0.0689 - val_mse: 0.0232 - val_root_mean_squared_error: 0.1522\n",
      "Epoch 14/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0145 - mae: 0.0145 - mse: 0.0022 - root_mean_squared_error: 0.0474 - val_loss: 0.0654 - val_mae: 0.0654 - val_mse: 0.0212 - val_root_mean_squared_error: 0.1456\n",
      "Epoch 15/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0143 - mae: 0.0143 - mse: 0.0022 - root_mean_squared_error: 0.0470 - val_loss: 0.0598 - val_mae: 0.0598 - val_mse: 0.0185 - val_root_mean_squared_error: 0.1360\n",
      "Epoch 16/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0141 - mae: 0.0141 - mse: 0.0022 - root_mean_squared_error: 0.0468 - val_loss: 0.0553 - val_mae: 0.0553 - val_mse: 0.0165 - val_root_mean_squared_error: 0.1285\n",
      "Epoch 17/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0139 - mae: 0.0139 - mse: 0.0022 - root_mean_squared_error: 0.0466 - val_loss: 0.0529 - val_mae: 0.0529 - val_mse: 0.0158 - val_root_mean_squared_error: 0.1258\n",
      "Epoch 18/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0138 - mae: 0.0138 - mse: 0.0022 - root_mean_squared_error: 0.0465 - val_loss: 0.0502 - val_mae: 0.0502 - val_mse: 0.0149 - val_root_mean_squared_error: 0.1219\n",
      "Epoch 19/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0136 - mae: 0.0136 - mse: 0.0021 - root_mean_squared_error: 0.0464 - val_loss: 0.0486 - val_mae: 0.0486 - val_mse: 0.0146 - val_root_mean_squared_error: 0.1207\n",
      "Epoch 20/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0134 - mae: 0.0134 - mse: 0.0021 - root_mean_squared_error: 0.0462 - val_loss: 0.0473 - val_mae: 0.0473 - val_mse: 0.0142 - val_root_mean_squared_error: 0.1193\n",
      "Epoch 21/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0133 - mae: 0.0133 - mse: 0.0021 - root_mean_squared_error: 0.0461 - val_loss: 0.0469 - val_mae: 0.0469 - val_mse: 0.0141 - val_root_mean_squared_error: 0.1188\n",
      "Epoch 22/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0131 - mae: 0.0131 - mse: 0.0021 - root_mean_squared_error: 0.0460 - val_loss: 0.0468 - val_mae: 0.0468 - val_mse: 0.0141 - val_root_mean_squared_error: 0.1188\n",
      "Epoch 23/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0130 - mae: 0.0130 - mse: 0.0021 - root_mean_squared_error: 0.0459 - val_loss: 0.0463 - val_mae: 0.0463 - val_mse: 0.0138 - val_root_mean_squared_error: 0.1176\n",
      "Epoch 24/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0129 - mae: 0.0129 - mse: 0.0021 - root_mean_squared_error: 0.0458 - val_loss: 0.0459 - val_mae: 0.0459 - val_mse: 0.0139 - val_root_mean_squared_error: 0.1177\n",
      "Epoch 25/300\n",
      "890/890 [==============================] - 47s 53ms/step - loss: 0.0127 - mae: 0.0127 - mse: 0.0021 - root_mean_squared_error: 0.0457 - val_loss: 0.0460 - val_mae: 0.0460 - val_mse: 0.0138 - val_root_mean_squared_error: 0.1174\n",
      "Epoch 26/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0126 - mae: 0.0126 - mse: 0.0021 - root_mean_squared_error: 0.0456 - val_loss: 0.0459 - val_mae: 0.0459 - val_mse: 0.0137 - val_root_mean_squared_error: 0.1170\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 27/300\n",
      "890/890 [==============================] - 50s 56ms/step - loss: 0.0124 - mae: 0.0124 - mse: 0.0021 - root_mean_squared_error: 0.0455 - val_loss: 0.0455 - val_mae: 0.0455 - val_mse: 0.0137 - val_root_mean_squared_error: 0.1170\n",
      "Epoch 28/300\n",
      "890/890 [==============================] - 48s 53ms/step - loss: 0.0122 - mae: 0.0122 - mse: 0.0021 - root_mean_squared_error: 0.0454 - val_loss: 0.0455 - val_mae: 0.0455 - val_mse: 0.0136 - val_root_mean_squared_error: 0.1166\n",
      "Epoch 29/300\n",
      "890/890 [==============================] - 47s 53ms/step - loss: 0.0121 - mae: 0.0121 - mse: 0.0021 - root_mean_squared_error: 0.0453 - val_loss: 0.0445 - val_mae: 0.0445 - val_mse: 0.0134 - val_root_mean_squared_error: 0.1158\n",
      "Epoch 30/300\n",
      "890/890 [==============================] - 47s 53ms/step - loss: 0.0119 - mae: 0.0119 - mse: 0.0020 - root_mean_squared_error: 0.0452 - val_loss: 0.0445 - val_mae: 0.0445 - val_mse: 0.0133 - val_root_mean_squared_error: 0.1151\n",
      "Epoch 31/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0118 - mae: 0.0118 - mse: 0.0020 - root_mean_squared_error: 0.0452 - val_loss: 0.0443 - val_mae: 0.0443 - val_mse: 0.0134 - val_root_mean_squared_error: 0.1158\n",
      "Epoch 32/300\n",
      "890/890 [==============================] - 47s 53ms/step - loss: 0.0118 - mae: 0.0118 - mse: 0.0020 - root_mean_squared_error: 0.0451 - val_loss: 0.0443 - val_mae: 0.0443 - val_mse: 0.0132 - val_root_mean_squared_error: 0.1151\n",
      "Epoch 33/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0117 - mae: 0.0117 - mse: 0.0020 - root_mean_squared_error: 0.0450 - val_loss: 0.0443 - val_mae: 0.0443 - val_mse: 0.0133 - val_root_mean_squared_error: 0.1152\n",
      "Epoch 34/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0116 - mae: 0.0116 - mse: 0.0020 - root_mean_squared_error: 0.0449 - val_loss: 0.0441 - val_mae: 0.0441 - val_mse: 0.0133 - val_root_mean_squared_error: 0.1151\n",
      "Epoch 35/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0116 - mae: 0.0116 - mse: 0.0020 - root_mean_squared_error: 0.0449 - val_loss: 0.0443 - val_mae: 0.0443 - val_mse: 0.0131 - val_root_mean_squared_error: 0.1145\n",
      "Epoch 36/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0115 - mae: 0.0115 - mse: 0.0020 - root_mean_squared_error: 0.0448 - val_loss: 0.0441 - val_mae: 0.0441 - val_mse: 0.0129 - val_root_mean_squared_error: 0.1137\n",
      "Epoch 37/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0114 - mae: 0.0114 - mse: 0.0020 - root_mean_squared_error: 0.0448 - val_loss: 0.0436 - val_mae: 0.0436 - val_mse: 0.0130 - val_root_mean_squared_error: 0.1139\n",
      "Epoch 38/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0114 - mae: 0.0114 - mse: 0.0020 - root_mean_squared_error: 0.0448 - val_loss: 0.0441 - val_mae: 0.0441 - val_mse: 0.0132 - val_root_mean_squared_error: 0.1150\n",
      "Epoch 39/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0113 - mae: 0.0113 - mse: 0.0020 - root_mean_squared_error: 0.0448 - val_loss: 0.0437 - val_mae: 0.0437 - val_mse: 0.0130 - val_root_mean_squared_error: 0.1140\n",
      "Epoch 40/300\n",
      "890/890 [==============================] - 49s 55ms/step - loss: 0.0113 - mae: 0.0113 - mse: 0.0020 - root_mean_squared_error: 0.0447 - val_loss: 0.0440 - val_mae: 0.0440 - val_mse: 0.0129 - val_root_mean_squared_error: 0.1137\n",
      "Epoch 41/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0112 - mae: 0.0112 - mse: 0.0020 - root_mean_squared_error: 0.0447 - val_loss: 0.0438 - val_mae: 0.0438 - val_mse: 0.0130 - val_root_mean_squared_error: 0.1140\n",
      "Epoch 42/300\n",
      "890/890 [==============================] - 48s 54ms/step - loss: 0.0112 - mae: 0.0112 - mse: 0.0020 - root_mean_squared_error: 0.0447 - val_loss: 0.0440 - val_mae: 0.0440 - val_mse: 0.0129 - val_root_mean_squared_error: 0.1137\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x21682016eb0>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from keras.layers import Input, Dense, Reshape, Flatten,Dropout\n",
    "from keras.models import Model\n",
    "from keras.optimizers import Adam\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "\n",
    "input_shape = (input_step,6,500)\n",
    "hidden_size1 = 256\n",
    "hidden_size2 = 128\n",
    "hidden_size3 = 100\n",
    "code_size = 64\n",
    "\n",
    "input_layer = Input(shape=(input_shape))\n",
    "print(input_layer)\n",
    "hidden_layer_0=Flatten()(input_layer)\n",
    "hidden_layer_1 = Dense(hidden_size1, activation='relu')(hidden_layer_0)\n",
    "hidden_layer_2 = Dense(hidden_size2, activation='relu')(hidden_layer_1)\n",
    "hidden_layer_3 = Dense(hidden_size3, activation='relu')(hidden_layer_2)\n",
    "bottleneck_layer = Dense(code_size*input_step, activation='sigmoid')(hidden_layer_3)\n",
    "\n",
    "# Define the decoder network\n",
    "hidden_layer_5 = Dense(hidden_size1, activation='relu')(bottleneck_layer)\n",
    "hidden_layer_6 = Dense(hidden_size2, activation='relu')(hidden_layer_5)\n",
    "hidden_layer_7 = Dense(hidden_size1, activation='relu')(hidden_layer_6)\n",
    "hidden_layer_8 = Dense(6*500*input_step, activation='sigmoid')(hidden_layer_7)\n",
    "output_layer=Reshape((input_step,6,500))(hidden_layer_8)\n",
    "\n",
    "autoencoder = Model(input_layer, output_layer)\n",
    "autoencoder.summary()\n",
    "opt = Adam(learning_rate=0.0001)\n",
    "autoencoder.compile(optimizer=opt, loss='mae',metrics=['mae', 'mse',tf.keras.metrics.RootMeanSquaredError()])\n",
    "early_stopping = EarlyStopping(monitor='val_loss', patience=5, mode='min')\n",
    "autoencoder.fit(X1, X1, epochs=300,batch_size=64,\n",
    "                validation_data=(X2, X2),callbacks=[early_stopping]\n",
    "              )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "a3cb8e3d",
   "metadata": {
    "id": "a3cb8e3d"
   },
   "outputs": [],
   "source": [
    "encoder = Model(input_layer, bottleneck_layer)\n",
    "encoder.compile(optimizer='adam', loss='mae')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "74201088",
   "metadata": {
    "id": "74201088",
    "outputId": "0cfb7f27-2da9-4a27-b445-d8472c454ffe"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1780/1780 [==============================] - 7s 4ms/step\n",
      "1719/1719 [==============================] - 7s 4ms/step\n"
     ]
    }
   ],
   "source": [
    "X11=encoder.predict(X1)\n",
    "X22=encoder.predict(X2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "7bdd3870",
   "metadata": {},
   "outputs": [],
   "source": [
    "decoder = Model(bottleneck_layer, output_layer)\n",
    "decoder.compile(optimizer='adam', loss='mae')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "8df2b56b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1780/1780 [==============================] - 8s 5ms/step\n",
      "1719/1719 [==============================] - 8s 5ms/step\n"
     ]
    }
   ],
   "source": [
    "Y1=decoder.predict(X11)\n",
    "Y2=decoder.predict(X22)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "13fc5b07",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x216825cf0a0>]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.plot(Y2[0][0][5])\n",
    "plt.plot(X2[0][0][5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "24558ad1",
   "metadata": {
    "id": "24558ad1"
   },
   "outputs": [],
   "source": [
    "np.save(\"X11_\"+str(input_step),X11)\n",
    "np.save(\"X22_\"+str(input_step),X22)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "dfbed698",
   "metadata": {
    "id": "dfbed698"
   },
   "outputs": [],
   "source": [
    "encoder.save(\"encoder-simple_\"+str(input_step)+\".h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "917775ff",
   "metadata": {},
   "source": [
    "# Cnn encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "f06b9808",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_7\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_6 (InputLayer)        [(None, 10, 6, 500)]      0         \n",
      "                                                                 \n",
      " conv2d_8 (Conv2D)           (None, 10, 6, 128)        576128    \n",
      "                                                                 \n",
      " max_pooling2d_2 (MaxPooling  (None, 10, 3, 128)       0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " conv2d_9 (Conv2D)           (None, 10, 3, 64)         73792     \n",
      "                                                                 \n",
      " max_pooling2d_3 (MaxPooling  (None, 10, 2, 64)        0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " conv2d_10 (Conv2D)          (None, 10, 2, 32)         18464     \n",
      "                                                                 \n",
      " reshape_7 (Reshape)         (None, 10, 1, 64)         0         \n",
      "                                                                 \n",
      " conv2d_11 (Conv2D)          (None, 10, 1, 64)         4160      \n",
      "                                                                 \n",
      " up_sampling2d_2 (UpSampling  (None, 10, 2, 64)        0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " conv2d_12 (Conv2D)          (None, 10, 2, 100)        6500      \n",
      "                                                                 \n",
      " up_sampling2d_3 (UpSampling  (None, 10, 4, 100)       0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " conv2d_13 (Conv2D)          (None, 10, 4, 128)        12928     \n",
      "                                                                 \n",
      " conv2d_14 (Conv2D)          (None, 10, 4, 252)        32508     \n",
      "                                                                 \n",
      " reshape_8 (Reshape)         (None, 10, 6, 168)        0         \n",
      "                                                                 \n",
      " dense_17 (Dense)            (None, 10, 6, 500)        84500     \n",
      "                                                                 \n",
      " reshape_9 (Reshape)         (None, 10, 6, 500)        0         \n",
      "                                                                 \n",
      " conv2d_15 (Conv2D)          (None, 10, 6, 500)        2250500   \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 3,059,480\n",
      "Trainable params: 3,059,480\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense, Reshape, UpSampling2D,Dropout\n",
    "from keras.models import Model\n",
    "\n",
    "# Define input shape\n",
    "input_shape = (input_step,6, 500)\n",
    "\n",
    "# Define encoder architecture\n",
    "inputs = Input(shape=input_shape)\n",
    "x = Conv2D(128,3, activation='relu', padding='same' )(inputs)\n",
    "\n",
    "x = MaxPooling2D((1,2), padding='same')(x)\n",
    "\n",
    "x = Conv2D(64,3, activation='relu', padding='same')(x)\n",
    "\n",
    "x = MaxPooling2D((1,2), padding='same')(x)\n",
    "x = Conv2D(32,3, activation='relu', padding='same')(x)\n",
    "\n",
    "encoded2 = Reshape((input_step,64))(x)\n",
    "\n",
    "\n",
    "\n",
    "# Define decoder architecture\n",
    "x = Reshape((input_step,1,64))(x)\n",
    "x = Conv2D(64, (1,1), activation='relu', padding='same')(x)\n",
    "x = UpSampling2D((1,2))(x)\n",
    "x = Conv2D(100, (1,1), activation='relu', padding='same')(x)\n",
    "x = UpSampling2D((1,2))(x)\n",
    "x = Conv2D(128, (1,1), activation='relu', padding='same')(x)\n",
    "x=Conv2D(252, (1,1), activation='sigmoid')(x)\n",
    "x=Reshape((input_step,6,168))(x)\n",
    "x=Dense(500, activation='linear')(x)\n",
    "x=Reshape((input_step,6,500))(x)\n",
    "\n",
    "decoded2 = Conv2D(500, 3, activation='sigmoid', padding='same')(x)\n",
    "\n",
    "# Define autoencoder model\n",
    "autoencoder2 = Model(inputs=inputs, outputs=decoded2)\n",
    "\n",
    "# Compile model\n",
    "autoencoder2.compile(optimizer='adam', loss='mae',metrics=['mae', 'mse',tf.keras.metrics.RootMeanSquaredError()])\n",
    "\n",
    "# Print model summary\n",
    "print(autoencoder2.summary())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "ccc956ef",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "890/890 [==============================] - 321s 359ms/step - loss: 0.0992 - mae: 0.0992 - mse: 0.0323 - root_mean_squared_error: 0.1798 - val_loss: 0.1280 - val_mae: 0.1280 - val_mse: 0.0464 - val_root_mean_squared_error: 0.2153\n",
      "Epoch 2/300\n",
      "890/890 [==============================] - 303s 341ms/step - loss: 0.0912 - mae: 0.0912 - mse: 0.0284 - root_mean_squared_error: 0.1686 - val_loss: 0.1191 - val_mae: 0.1191 - val_mse: 0.0437 - val_root_mean_squared_error: 0.2091\n",
      "Epoch 3/300\n",
      "890/890 [==============================] - 293s 329ms/step - loss: 0.0889 - mae: 0.0889 - mse: 0.0277 - root_mean_squared_error: 0.1666 - val_loss: 0.1181 - val_mae: 0.1181 - val_mse: 0.0433 - val_root_mean_squared_error: 0.2082\n",
      "Epoch 4/300\n",
      "890/890 [==============================] - 310s 348ms/step - loss: 0.0878 - mae: 0.0878 - mse: 0.0273 - root_mean_squared_error: 0.1653 - val_loss: 0.1171 - val_mae: 0.1171 - val_mse: 0.0422 - val_root_mean_squared_error: 0.2055\n",
      "Epoch 5/300\n",
      "890/890 [==============================] - 306s 344ms/step - loss: 0.0870 - mae: 0.0870 - mse: 0.0273 - root_mean_squared_error: 0.1651 - val_loss: 0.1165 - val_mae: 0.1165 - val_mse: 0.0418 - val_root_mean_squared_error: 0.2044\n",
      "Epoch 6/300\n",
      "890/890 [==============================] - 306s 343ms/step - loss: 0.0865 - mae: 0.0865 - mse: 0.0272 - root_mean_squared_error: 0.1650 - val_loss: 0.1147 - val_mae: 0.1147 - val_mse: 0.0412 - val_root_mean_squared_error: 0.2029\n",
      "Epoch 7/300\n",
      "890/890 [==============================] - 295s 332ms/step - loss: 0.0860 - mae: 0.0860 - mse: 0.0272 - root_mean_squared_error: 0.1649 - val_loss: 0.1151 - val_mae: 0.1151 - val_mse: 0.0413 - val_root_mean_squared_error: 0.2032\n",
      "Epoch 8/300\n",
      "890/890 [==============================] - 289s 325ms/step - loss: 0.0856 - mae: 0.0856 - mse: 0.0271 - root_mean_squared_error: 0.1646 - val_loss: 0.1155 - val_mae: 0.1155 - val_mse: 0.0411 - val_root_mean_squared_error: 0.2027\n",
      "Epoch 9/300\n",
      "890/890 [==============================] - 294s 331ms/step - loss: 0.0854 - mae: 0.0854 - mse: 0.0272 - root_mean_squared_error: 0.1648 - val_loss: 0.1142 - val_mae: 0.1142 - val_mse: 0.0410 - val_root_mean_squared_error: 0.2024\n",
      "Epoch 10/300\n",
      "890/890 [==============================] - 310s 348ms/step - loss: 0.0851 - mae: 0.0851 - mse: 0.0272 - root_mean_squared_error: 0.1648 - val_loss: 0.1143 - val_mae: 0.1143 - val_mse: 0.0411 - val_root_mean_squared_error: 0.2027\n",
      "Epoch 11/300\n",
      "890/890 [==============================] - 293s 329ms/step - loss: 0.0849 - mae: 0.0849 - mse: 0.0272 - root_mean_squared_error: 0.1649 - val_loss: 0.1153 - val_mae: 0.1153 - val_mse: 0.0412 - val_root_mean_squared_error: 0.2030\n",
      "Epoch 12/300\n",
      "890/890 [==============================] - 286s 322ms/step - loss: 0.0848 - mae: 0.0848 - mse: 0.0273 - root_mean_squared_error: 0.1652 - val_loss: 0.1152 - val_mae: 0.1152 - val_mse: 0.0421 - val_root_mean_squared_error: 0.2051\n",
      "Epoch 13/300\n",
      "890/890 [==============================] - 247s 277ms/step - loss: 0.0846 - mae: 0.0846 - mse: 0.0273 - root_mean_squared_error: 0.1652 - val_loss: 0.1138 - val_mae: 0.1138 - val_mse: 0.0417 - val_root_mean_squared_error: 0.2041\n",
      "Epoch 14/300\n",
      "890/890 [==============================] - 244s 274ms/step - loss: 0.0844 - mae: 0.0844 - mse: 0.0272 - root_mean_squared_error: 0.1650 - val_loss: 0.1133 - val_mae: 0.1133 - val_mse: 0.0417 - val_root_mean_squared_error: 0.2041\n",
      "Epoch 15/300\n",
      "890/890 [==============================] - 244s 274ms/step - loss: 0.0843 - mae: 0.0843 - mse: 0.0273 - root_mean_squared_error: 0.1652 - val_loss: 0.1121 - val_mae: 0.1121 - val_mse: 0.0408 - val_root_mean_squared_error: 0.2020\n",
      "Epoch 16/300\n",
      "890/890 [==============================] - 251s 283ms/step - loss: 0.0841 - mae: 0.0841 - mse: 0.0273 - root_mean_squared_error: 0.1651 - val_loss: 0.1110 - val_mae: 0.1110 - val_mse: 0.0408 - val_root_mean_squared_error: 0.2020\n",
      "Epoch 17/300\n",
      "890/890 [==============================] - 259s 291ms/step - loss: 0.0840 - mae: 0.0840 - mse: 0.0272 - root_mean_squared_error: 0.1648 - val_loss: 0.1097 - val_mae: 0.1097 - val_mse: 0.0403 - val_root_mean_squared_error: 0.2008\n",
      "Epoch 18/300\n",
      "890/890 [==============================] - 272s 306ms/step - loss: 0.0839 - mae: 0.0839 - mse: 0.0273 - root_mean_squared_error: 0.1651 - val_loss: 0.1098 - val_mae: 0.1098 - val_mse: 0.0404 - val_root_mean_squared_error: 0.2009\n",
      "Epoch 19/300\n",
      "890/890 [==============================] - 303s 341ms/step - loss: 0.0838 - mae: 0.0838 - mse: 0.0273 - root_mean_squared_error: 0.1652 - val_loss: 0.1096 - val_mae: 0.1096 - val_mse: 0.0403 - val_root_mean_squared_error: 0.2008\n",
      "Epoch 20/300\n",
      "890/890 [==============================] - 274s 308ms/step - loss: 0.0838 - mae: 0.0838 - mse: 0.0273 - root_mean_squared_error: 0.1652 - val_loss: 0.1097 - val_mae: 0.1097 - val_mse: 0.0406 - val_root_mean_squared_error: 0.2014\n",
      "Epoch 21/300\n",
      "890/890 [==============================] - 265s 298ms/step - loss: 0.0837 - mae: 0.0837 - mse: 0.0274 - root_mean_squared_error: 0.1654 - val_loss: 0.1089 - val_mae: 0.1089 - val_mse: 0.0404 - val_root_mean_squared_error: 0.2010\n",
      "Epoch 22/300\n",
      "890/890 [==============================] - 266s 299ms/step - loss: 0.0836 - mae: 0.0836 - mse: 0.0274 - root_mean_squared_error: 0.1655 - val_loss: 0.1089 - val_mae: 0.1089 - val_mse: 0.0404 - val_root_mean_squared_error: 0.2011\n",
      "Epoch 23/300\n",
      "890/890 [==============================] - 264s 297ms/step - loss: 0.0835 - mae: 0.0835 - mse: 0.0274 - root_mean_squared_error: 0.1655 - val_loss: 0.1102 - val_mae: 0.1102 - val_mse: 0.0409 - val_root_mean_squared_error: 0.2022\n",
      "Epoch 24/300\n",
      "890/890 [==============================] - 264s 297ms/step - loss: 0.0834 - mae: 0.0834 - mse: 0.0275 - root_mean_squared_error: 0.1657 - val_loss: 0.1080 - val_mae: 0.1080 - val_mse: 0.0402 - val_root_mean_squared_error: 0.2005\n",
      "Epoch 25/300\n",
      "890/890 [==============================] - 264s 297ms/step - loss: 0.0834 - mae: 0.0834 - mse: 0.0275 - root_mean_squared_error: 0.1659 - val_loss: 0.1085 - val_mae: 0.1085 - val_mse: 0.0401 - val_root_mean_squared_error: 0.2001\n",
      "Epoch 26/300\n",
      "890/890 [==============================] - 265s 298ms/step - loss: 0.0833 - mae: 0.0833 - mse: 0.0275 - root_mean_squared_error: 0.1658 - val_loss: 0.1087 - val_mae: 0.1087 - val_mse: 0.0395 - val_root_mean_squared_error: 0.1988\n",
      "Epoch 27/300\n",
      "890/890 [==============================] - 261s 293ms/step - loss: 0.0832 - mae: 0.0832 - mse: 0.0275 - root_mean_squared_error: 0.1658 - val_loss: 0.1074 - val_mae: 0.1074 - val_mse: 0.0395 - val_root_mean_squared_error: 0.1988\n",
      "Epoch 28/300\n",
      "890/890 [==============================] - 260s 292ms/step - loss: 0.0832 - mae: 0.0832 - mse: 0.0275 - root_mean_squared_error: 0.1660 - val_loss: 0.1064 - val_mae: 0.1064 - val_mse: 0.0398 - val_root_mean_squared_error: 0.1996\n",
      "Epoch 29/300\n",
      "890/890 [==============================] - 255s 286ms/step - loss: 0.0831 - mae: 0.0831 - mse: 0.0276 - root_mean_squared_error: 0.1660 - val_loss: 0.1088 - val_mae: 0.1088 - val_mse: 0.0398 - val_root_mean_squared_error: 0.1994\n",
      "Epoch 30/300\n",
      "890/890 [==============================] - 256s 288ms/step - loss: 0.0831 - mae: 0.0831 - mse: 0.0275 - root_mean_squared_error: 0.1659 - val_loss: 0.1080 - val_mae: 0.1080 - val_mse: 0.0403 - val_root_mean_squared_error: 0.2007\n",
      "Epoch 31/300\n",
      "890/890 [==============================] - 265s 297ms/step - loss: 0.0830 - mae: 0.0830 - mse: 0.0276 - root_mean_squared_error: 0.1661 - val_loss: 0.1080 - val_mae: 0.1080 - val_mse: 0.0400 - val_root_mean_squared_error: 0.1999\n",
      "Epoch 32/300\n",
      "890/890 [==============================] - 260s 293ms/step - loss: 0.0829 - mae: 0.0829 - mse: 0.0276 - root_mean_squared_error: 0.1662 - val_loss: 0.1094 - val_mae: 0.1094 - val_mse: 0.0408 - val_root_mean_squared_error: 0.2019\n",
      "Epoch 33/300\n",
      "890/890 [==============================] - 251s 282ms/step - loss: 0.0829 - mae: 0.0829 - mse: 0.0276 - root_mean_squared_error: 0.1662 - val_loss: 0.1092 - val_mae: 0.1092 - val_mse: 0.0410 - val_root_mean_squared_error: 0.2026\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x21682651f10>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train the model\n",
    "opt = Adam(learning_rate=0.0001)\n",
    "autoencoder.compile(optimizer=opt, loss='mae',metrics=['mae', 'mse',tf.keras.metrics.RootMeanSquaredError()])\n",
    "early_stopping = EarlyStopping(monitor='val_loss', patience=5, mode='min')\n",
    "autoencoder2.fit(X1, X1, epochs=300,batch_size=64,\n",
    "                validation_data=(X2, X2),callbacks=[early_stopping]\n",
    "              )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "a59812f1",
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_2 = Model(inputs, encoded2)\n",
    "encoder_2.compile(optimizer='adam', loss='mae')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "85f75f9d",
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_2.save(\"encoder_cnn_\"+str(input_step)+\".h5\")"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}