9-3_IMDB_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "AlaMdWF26k2H"
   },
   "source": [
    "# LSTM과 GRU 셀"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "o_Csb7v96k2M"
   },
   "source": [
    "<table align=\"left\">\n",
    "  <td>\n",
    "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/rickiepark/hg-mldl/blob/master/9-3.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />구글 코랩에서 실행하기</a>\n",
    "  </td>\n",
    "</table>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KWi9yl0q6k2N"
   },
   "source": [
    "## LSTM 신경망 훈련하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:32:03.992822Z",
     "iopub.status.busy": "2022-01-08T06:32:03.992104Z",
     "iopub.status.idle": "2022-01-08T06:32:10.411204Z",
     "shell.execute_reply": "2022-01-08T06:32:10.410270Z"
    },
    "id": "XL7x0NCx6k2N",
    "outputId": "d5c9f55e-6a51-49b7-e739-d2242adddcae"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.datasets import imdb\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "(train_input, train_target), (test_input, test_target) = imdb.load_data(\n",
    "    num_words=500)\n",
    "\n",
    "train_input, val_input, train_target, val_target = train_test_split(\n",
    "    train_input, train_target, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-01-08T06:32:10.420589Z",
     "iopub.status.busy": "2022-01-08T06:32:10.416057Z",
     "iopub.status.idle": "2022-01-08T06:32:10.617660Z",
     "shell.execute_reply": "2022-01-08T06:32:10.618191Z"
    },
    "id": "5bjJ14QJ6k2O"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "train_seq = pad_sequences(train_input, maxlen=100)\n",
    "val_seq = pad_sequences(val_input, maxlen=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:32:10.645231Z",
     "iopub.status.busy": "2022-01-08T06:32:10.623014Z",
     "iopub.status.idle": "2022-01-08T06:32:10.923445Z",
     "shell.execute_reply": "2022-01-08T06:32:10.922126Z"
    },
    "id": "uMlTD4zv6k2O",
    "outputId": "dc7c0080-f72c-4032-caf4-1a8e63202989"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-01-08 06:32:10.646756: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 100, 16)           8000      \n",
      "                                                                 \n",
      " lstm (LSTM)                 (None, 8)                 800       \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 9         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 8,809\n",
      "Trainable params: 8,809\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow import keras\n",
    "\n",
    "model = keras.Sequential()\n",
    "\n",
    "model.add(keras.layers.Embedding(500, 16, input_length=100))\n",
    "model.add(keras.layers.LSTM(8))\n",
    "model.add(keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:32:10.933595Z",
     "iopub.status.busy": "2022-01-08T06:32:10.926955Z",
     "iopub.status.idle": "2022-01-08T06:40:41.719394Z",
     "shell.execute_reply": "2022-01-08T06:40:41.718536Z"
    },
    "id": "nCmqOkjI6k2P",
    "outputId": "5057828c-5f90-48a5-fafa-c31f7407cc4c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "313/313 [==============================] - 14s 40ms/step - loss: 0.6923 - accuracy: 0.5464 - val_loss: 0.6913 - val_accuracy: 0.5820\n",
      "Epoch 2/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.6897 - accuracy: 0.6051 - val_loss: 0.6879 - val_accuracy: 0.6180\n",
      "Epoch 3/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.6836 - accuracy: 0.6439 - val_loss: 0.6787 - val_accuracy: 0.6432\n",
      "Epoch 4/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.6595 - accuracy: 0.6526 - val_loss: 0.6346 - val_accuracy: 0.6884\n",
      "Epoch 5/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5967 - accuracy: 0.7280 - val_loss: 0.5802 - val_accuracy: 0.7412\n",
      "Epoch 6/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.5594 - accuracy: 0.7525 - val_loss: 0.5513 - val_accuracy: 0.7428\n",
      "Epoch 7/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5335 - accuracy: 0.7606 - val_loss: 0.5271 - val_accuracy: 0.7570\n",
      "Epoch 8/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.5118 - accuracy: 0.7707 - val_loss: 0.5102 - val_accuracy: 0.7680\n",
      "Epoch 9/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4951 - accuracy: 0.7802 - val_loss: 0.4954 - val_accuracy: 0.7762\n",
      "Epoch 10/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4815 - accuracy: 0.7869 - val_loss: 0.4878 - val_accuracy: 0.7748\n",
      "Epoch 11/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4711 - accuracy: 0.7928 - val_loss: 0.4798 - val_accuracy: 0.7798\n",
      "Epoch 12/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4621 - accuracy: 0.7979 - val_loss: 0.4730 - val_accuracy: 0.7820\n",
      "Epoch 13/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4550 - accuracy: 0.8009 - val_loss: 0.4659 - val_accuracy: 0.7894\n",
      "Epoch 14/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4492 - accuracy: 0.8030 - val_loss: 0.4614 - val_accuracy: 0.7908\n",
      "Epoch 15/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4440 - accuracy: 0.8044 - val_loss: 0.4593 - val_accuracy: 0.7902\n",
      "Epoch 16/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4394 - accuracy: 0.8051 - val_loss: 0.4562 - val_accuracy: 0.7922\n",
      "Epoch 17/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4362 - accuracy: 0.8058 - val_loss: 0.4526 - val_accuracy: 0.7926\n",
      "Epoch 18/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4329 - accuracy: 0.8089 - val_loss: 0.4514 - val_accuracy: 0.7932\n",
      "Epoch 19/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4300 - accuracy: 0.8113 - val_loss: 0.4481 - val_accuracy: 0.7966\n",
      "Epoch 20/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4275 - accuracy: 0.8101 - val_loss: 0.4462 - val_accuracy: 0.7970\n",
      "Epoch 21/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4252 - accuracy: 0.8125 - val_loss: 0.4450 - val_accuracy: 0.7988\n",
      "Epoch 22/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4233 - accuracy: 0.8137 - val_loss: 0.4504 - val_accuracy: 0.7868\n",
      "Epoch 23/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4222 - accuracy: 0.8131 - val_loss: 0.4435 - val_accuracy: 0.7992\n",
      "Epoch 24/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4201 - accuracy: 0.8131 - val_loss: 0.4431 - val_accuracy: 0.7958\n",
      "Epoch 25/100\n",
      "313/313 [==============================] - 11s 37ms/step - loss: 0.4194 - accuracy: 0.8145 - val_loss: 0.4410 - val_accuracy: 0.7990\n",
      "Epoch 26/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4178 - accuracy: 0.8151 - val_loss: 0.4409 - val_accuracy: 0.7972\n",
      "Epoch 27/100\n",
      "313/313 [==============================] - 11s 37ms/step - loss: 0.4170 - accuracy: 0.8141 - val_loss: 0.4379 - val_accuracy: 0.7998\n",
      "Epoch 28/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4155 - accuracy: 0.8145 - val_loss: 0.4372 - val_accuracy: 0.8000\n",
      "Epoch 29/100\n",
      "313/313 [==============================] - 11s 36ms/step - loss: 0.4150 - accuracy: 0.8159 - val_loss: 0.4401 - val_accuracy: 0.7956\n",
      "Epoch 30/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4141 - accuracy: 0.8159 - val_loss: 0.4375 - val_accuracy: 0.8002\n",
      "Epoch 31/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4132 - accuracy: 0.8156 - val_loss: 0.4351 - val_accuracy: 0.8012\n",
      "Epoch 32/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4126 - accuracy: 0.8141 - val_loss: 0.4366 - val_accuracy: 0.7992\n",
      "Epoch 33/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4118 - accuracy: 0.8160 - val_loss: 0.4348 - val_accuracy: 0.7976\n",
      "Epoch 34/100\n",
      "313/313 [==============================] - 11s 37ms/step - loss: 0.4109 - accuracy: 0.8159 - val_loss: 0.4340 - val_accuracy: 0.7996\n",
      "Epoch 35/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4103 - accuracy: 0.8155 - val_loss: 0.4332 - val_accuracy: 0.8018\n",
      "Epoch 36/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4097 - accuracy: 0.8161 - val_loss: 0.4345 - val_accuracy: 0.7992\n",
      "Epoch 37/100\n",
      "313/313 [==============================] - 11s 36ms/step - loss: 0.4091 - accuracy: 0.8157 - val_loss: 0.4329 - val_accuracy: 0.8030\n",
      "Epoch 38/100\n",
      "313/313 [==============================] - 11s 37ms/step - loss: 0.4086 - accuracy: 0.8143 - val_loss: 0.4317 - val_accuracy: 0.8046\n",
      "Epoch 39/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4080 - accuracy: 0.8161 - val_loss: 0.4343 - val_accuracy: 0.8002\n",
      "Epoch 40/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4077 - accuracy: 0.8164 - val_loss: 0.4311 - val_accuracy: 0.8030\n",
      "Epoch 41/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4070 - accuracy: 0.8160 - val_loss: 0.4311 - val_accuracy: 0.8022\n",
      "Epoch 42/100\n",
      "313/313 [==============================] - 12s 37ms/step - loss: 0.4062 - accuracy: 0.8181 - val_loss: 0.4340 - val_accuracy: 0.8010\n",
      "Epoch 43/100\n",
      "313/313 [==============================] - 11s 37ms/step - loss: 0.4058 - accuracy: 0.8163 - val_loss: 0.4321 - val_accuracy: 0.8032\n"
     ]
    }
   ],
   "source": [
    "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n",
    "model.compile(optimizer=rmsprop, loss='binary_crossentropy', \n",
    "              metrics=['accuracy'])\n",
    "\n",
    "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-lstm-model.h5', \n",
    "                                                save_best_only=True)\n",
    "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n",
    "                                                  restore_best_weights=True)\n",
    "\n",
    "history = model.fit(train_seq, train_target, epochs=100, batch_size=64,\n",
    "                    validation_data=(val_seq, val_target),\n",
    "                    callbacks=[checkpoint_cb, early_stopping_cb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:40:41.727198Z",
     "iopub.status.busy": "2022-01-08T06:40:41.724610Z",
     "iopub.status.idle": "2022-01-08T06:40:42.690300Z",
     "shell.execute_reply": "2022-01-08T06:40:42.689672Z"
    },
    "id": "Xl1bliNr6k2Q",
    "outputId": "4e6792a9-2118-4498-a4cb-f0b39d1609c0"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train', 'val'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ob79h7Kc6k2Q"
   },
   "source": [
    "## 순환 층에 드롭아웃 적용하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-01-08T06:40:42.697123Z",
     "iopub.status.busy": "2022-01-08T06:40:42.696482Z",
     "iopub.status.idle": "2022-01-08T06:40:42.919396Z",
     "shell.execute_reply": "2022-01-08T06:40:42.918675Z"
    },
    "id": "Fr0DUR-Y6k2R"
   },
   "outputs": [],
   "source": [
    "model2 = keras.Sequential()\n",
    "\n",
    "model2.add(keras.layers.Embedding(500, 16, input_length=100))\n",
    "model2.add(keras.layers.LSTM(8, dropout=0.3))\n",
    "model2.add(keras.layers.Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:40:42.932923Z",
     "iopub.status.busy": "2022-01-08T06:40:42.932101Z",
     "iopub.status.idle": "2022-01-08T06:50:43.260633Z",
     "shell.execute_reply": "2022-01-08T06:50:43.259967Z"
    },
    "id": "Ap8XxUO46k2R",
    "outputId": "5953909f-5c48-43e9-b321-8b5f3ed0d52a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "313/313 [==============================] - 14s 40ms/step - loss: 0.6923 - accuracy: 0.5468 - val_loss: 0.6912 - val_accuracy: 0.6046\n",
      "Epoch 2/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.6893 - accuracy: 0.6141 - val_loss: 0.6861 - val_accuracy: 0.6644\n",
      "Epoch 3/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.6778 - accuracy: 0.6650 - val_loss: 0.6600 - val_accuracy: 0.7026\n",
      "Epoch 4/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.6167 - accuracy: 0.7024 - val_loss: 0.5843 - val_accuracy: 0.7162\n",
      "Epoch 5/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5748 - accuracy: 0.7236 - val_loss: 0.5587 - val_accuracy: 0.7398\n",
      "Epoch 6/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5519 - accuracy: 0.7398 - val_loss: 0.5390 - val_accuracy: 0.7512\n",
      "Epoch 7/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.5328 - accuracy: 0.7549 - val_loss: 0.5242 - val_accuracy: 0.7600\n",
      "Epoch 8/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5149 - accuracy: 0.7667 - val_loss: 0.5069 - val_accuracy: 0.7700\n",
      "Epoch 9/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.5009 - accuracy: 0.7742 - val_loss: 0.4946 - val_accuracy: 0.7752\n",
      "Epoch 10/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4864 - accuracy: 0.7821 - val_loss: 0.4829 - val_accuracy: 0.7856\n",
      "Epoch 11/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4760 - accuracy: 0.7872 - val_loss: 0.4737 - val_accuracy: 0.7888\n",
      "Epoch 12/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4687 - accuracy: 0.7908 - val_loss: 0.4678 - val_accuracy: 0.7920\n",
      "Epoch 13/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4614 - accuracy: 0.7932 - val_loss: 0.4614 - val_accuracy: 0.7926\n",
      "Epoch 14/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4532 - accuracy: 0.7986 - val_loss: 0.4565 - val_accuracy: 0.7960\n",
      "Epoch 15/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4500 - accuracy: 0.7990 - val_loss: 0.4525 - val_accuracy: 0.7962\n",
      "Epoch 16/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4452 - accuracy: 0.8020 - val_loss: 0.4492 - val_accuracy: 0.7972\n",
      "Epoch 17/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4396 - accuracy: 0.8044 - val_loss: 0.4472 - val_accuracy: 0.7958\n",
      "Epoch 18/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4375 - accuracy: 0.8030 - val_loss: 0.4446 - val_accuracy: 0.7956\n",
      "Epoch 19/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4337 - accuracy: 0.8031 - val_loss: 0.4418 - val_accuracy: 0.7978\n",
      "Epoch 20/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4321 - accuracy: 0.8047 - val_loss: 0.4393 - val_accuracy: 0.7980\n",
      "Epoch 21/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4291 - accuracy: 0.8060 - val_loss: 0.4380 - val_accuracy: 0.8000\n",
      "Epoch 22/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4271 - accuracy: 0.8070 - val_loss: 0.4371 - val_accuracy: 0.7998\n",
      "Epoch 23/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4259 - accuracy: 0.8073 - val_loss: 0.4362 - val_accuracy: 0.7980\n",
      "Epoch 24/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4224 - accuracy: 0.8070 - val_loss: 0.4375 - val_accuracy: 0.8000\n",
      "Epoch 25/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4246 - accuracy: 0.8066 - val_loss: 0.4349 - val_accuracy: 0.7956\n",
      "Epoch 26/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4210 - accuracy: 0.8064 - val_loss: 0.4326 - val_accuracy: 0.8000\n",
      "Epoch 27/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4206 - accuracy: 0.8078 - val_loss: 0.4317 - val_accuracy: 0.7990\n",
      "Epoch 28/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4181 - accuracy: 0.8079 - val_loss: 0.4303 - val_accuracy: 0.8018\n",
      "Epoch 29/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4175 - accuracy: 0.8092 - val_loss: 0.4305 - val_accuracy: 0.8002\n",
      "Epoch 30/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4177 - accuracy: 0.8090 - val_loss: 0.4308 - val_accuracy: 0.7994\n",
      "Epoch 31/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4167 - accuracy: 0.8101 - val_loss: 0.4293 - val_accuracy: 0.8024\n",
      "Epoch 32/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4154 - accuracy: 0.8103 - val_loss: 0.4291 - val_accuracy: 0.8008\n",
      "Epoch 33/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4151 - accuracy: 0.8100 - val_loss: 0.4284 - val_accuracy: 0.7996\n",
      "Epoch 34/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4123 - accuracy: 0.8101 - val_loss: 0.4273 - val_accuracy: 0.8018\n",
      "Epoch 35/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4136 - accuracy: 0.8080 - val_loss: 0.4279 - val_accuracy: 0.8024\n",
      "Epoch 36/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4097 - accuracy: 0.8151 - val_loss: 0.4269 - val_accuracy: 0.8026\n",
      "Epoch 37/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4115 - accuracy: 0.8140 - val_loss: 0.4269 - val_accuracy: 0.8036\n",
      "Epoch 38/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4099 - accuracy: 0.8120 - val_loss: 0.4264 - val_accuracy: 0.8032\n",
      "Epoch 39/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4106 - accuracy: 0.8127 - val_loss: 0.4260 - val_accuracy: 0.8022\n",
      "Epoch 40/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4088 - accuracy: 0.8118 - val_loss: 0.4295 - val_accuracy: 0.8014\n",
      "Epoch 41/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4067 - accuracy: 0.8151 - val_loss: 0.4257 - val_accuracy: 0.8040\n",
      "Epoch 42/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4067 - accuracy: 0.8128 - val_loss: 0.4268 - val_accuracy: 0.8020\n",
      "Epoch 43/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4059 - accuracy: 0.8141 - val_loss: 0.4253 - val_accuracy: 0.8030\n",
      "Epoch 44/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4054 - accuracy: 0.8151 - val_loss: 0.4307 - val_accuracy: 0.8014\n",
      "Epoch 45/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4047 - accuracy: 0.8163 - val_loss: 0.4252 - val_accuracy: 0.8034\n",
      "Epoch 46/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4062 - accuracy: 0.8130 - val_loss: 0.4247 - val_accuracy: 0.8044\n",
      "Epoch 47/100\n",
      "313/313 [==============================] - 12s 40ms/step - loss: 0.4056 - accuracy: 0.8134 - val_loss: 0.4255 - val_accuracy: 0.8012\n",
      "Epoch 48/100\n",
      "313/313 [==============================] - 12s 39ms/step - loss: 0.4038 - accuracy: 0.8161 - val_loss: 0.4251 - val_accuracy: 0.8030\n",
      "Epoch 49/100\n",
      "313/313 [==============================] - 12s 38ms/step - loss: 0.4032 - accuracy: 0.8147 - val_loss: 0.4270 - val_accuracy: 0.8012\n"
     ]
    }
   ],
   "source": [
    "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n",
    "model2.compile(optimizer=rmsprop, loss='binary_crossentropy', \n",
    "               metrics=['accuracy'])\n",
    "\n",
    "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-dropout-model.h5', \n",
    "                                                save_best_only=True)\n",
    "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n",
    "                                                  restore_best_weights=True)\n",
    "\n",
    "history = model2.fit(train_seq, train_target, epochs=100, batch_size=64,\n",
    "                     validation_data=(val_seq, val_target),\n",
    "                     callbacks=[checkpoint_cb, early_stopping_cb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:50:43.264141Z",
     "iopub.status.busy": "2022-01-08T06:50:43.263215Z",
     "iopub.status.idle": "2022-01-08T06:50:43.413447Z",
     "shell.execute_reply": "2022-01-08T06:50:43.412882Z"
    },
    "id": "Qzwsp3pl6k2R",
    "outputId": "97305973-3c11-4c55-c466-a608273094cb"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train', 'val'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DKtQByM96k2S"
   },
   "source": [
    "## 2개의 층을 연결하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:50:43.420502Z",
     "iopub.status.busy": "2022-01-08T06:50:43.419839Z",
     "iopub.status.idle": "2022-01-08T06:50:43.869743Z",
     "shell.execute_reply": "2022-01-08T06:50:43.869180Z"
    },
    "id": "YkQboCpD6k2S",
    "outputId": "5f30a5ec-16a9-476a-95b6-3d55b36b64b4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_2 (Embedding)     (None, 100, 16)           8000      \n",
      "                                                                 \n",
      " lstm_2 (LSTM)               (None, 100, 8)            800       \n",
      "                                                                 \n",
      " lstm_3 (LSTM)               (None, 8)                 544       \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 1)                 9         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 9,353\n",
      "Trainable params: 9,353\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model3 = keras.Sequential()\n",
    "\n",
    "model3.add(keras.layers.Embedding(500, 16, input_length=100))\n",
    "model3.add(keras.layers.LSTM(8, dropout=0.3, return_sequences=True))\n",
    "model3.add(keras.layers.LSTM(8, dropout=0.3))\n",
    "model3.add(keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "model3.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T06:50:43.878202Z",
     "iopub.status.busy": "2022-01-08T06:50:43.877451Z",
     "iopub.status.idle": "2022-01-08T07:03:47.427095Z",
     "shell.execute_reply": "2022-01-08T07:03:47.426473Z"
    },
    "id": "hDk_ffAo6k2S",
    "outputId": "f7439e48-073f-4124-b10e-dea60f18f69d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "313/313 [==============================] - 28s 77ms/step - loss: 0.6926 - accuracy: 0.5441 - val_loss: 0.6918 - val_accuracy: 0.5796\n",
      "Epoch 2/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.6892 - accuracy: 0.5978 - val_loss: 0.6856 - val_accuracy: 0.6170\n",
      "Epoch 3/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.6687 - accuracy: 0.6461 - val_loss: 0.6383 - val_accuracy: 0.6836\n",
      "Epoch 4/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.5917 - accuracy: 0.7092 - val_loss: 0.5636 - val_accuracy: 0.7230\n",
      "Epoch 5/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.5468 - accuracy: 0.7393 - val_loss: 0.5349 - val_accuracy: 0.7456\n",
      "Epoch 6/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.5279 - accuracy: 0.7466 - val_loss: 0.5159 - val_accuracy: 0.7574\n",
      "Epoch 7/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.5141 - accuracy: 0.7578 - val_loss: 0.5025 - val_accuracy: 0.7664\n",
      "Epoch 8/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.5008 - accuracy: 0.7646 - val_loss: 0.4909 - val_accuracy: 0.7710\n",
      "Epoch 9/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.4883 - accuracy: 0.7752 - val_loss: 0.4842 - val_accuracy: 0.7712\n",
      "Epoch 10/100\n",
      "313/313 [==============================] - 23s 74ms/step - loss: 0.4805 - accuracy: 0.7796 - val_loss: 0.4771 - val_accuracy: 0.7754\n",
      "Epoch 11/100\n",
      "313/313 [==============================] - 24s 75ms/step - loss: 0.4765 - accuracy: 0.7801 - val_loss: 0.4715 - val_accuracy: 0.7790\n",
      "Epoch 12/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.4695 - accuracy: 0.7842 - val_loss: 0.4669 - val_accuracy: 0.7820\n",
      "Epoch 13/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.4669 - accuracy: 0.7847 - val_loss: 0.4630 - val_accuracy: 0.7806\n",
      "Epoch 14/100\n",
      "313/313 [==============================] - 24s 75ms/step - loss: 0.4611 - accuracy: 0.7886 - val_loss: 0.4592 - val_accuracy: 0.7860\n",
      "Epoch 15/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.4568 - accuracy: 0.7914 - val_loss: 0.4588 - val_accuracy: 0.7836\n",
      "Epoch 16/100\n",
      "313/313 [==============================] - 24s 75ms/step - loss: 0.4529 - accuracy: 0.7910 - val_loss: 0.4577 - val_accuracy: 0.7874\n",
      "Epoch 17/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4515 - accuracy: 0.7948 - val_loss: 0.4532 - val_accuracy: 0.7862\n",
      "Epoch 18/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4478 - accuracy: 0.7952 - val_loss: 0.4547 - val_accuracy: 0.7864\n",
      "Epoch 19/100\n",
      "313/313 [==============================] - 24s 77ms/step - loss: 0.4446 - accuracy: 0.7961 - val_loss: 0.4495 - val_accuracy: 0.7898\n",
      "Epoch 20/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4428 - accuracy: 0.7980 - val_loss: 0.4476 - val_accuracy: 0.7930\n",
      "Epoch 21/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4400 - accuracy: 0.7995 - val_loss: 0.4460 - val_accuracy: 0.7920\n",
      "Epoch 22/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4410 - accuracy: 0.7964 - val_loss: 0.4503 - val_accuracy: 0.7884\n",
      "Epoch 23/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4366 - accuracy: 0.7993 - val_loss: 0.4510 - val_accuracy: 0.7910\n",
      "Epoch 24/100\n",
      "313/313 [==============================] - 24s 77ms/step - loss: 0.4344 - accuracy: 0.8027 - val_loss: 0.4420 - val_accuracy: 0.7968\n",
      "Epoch 25/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4340 - accuracy: 0.8012 - val_loss: 0.4416 - val_accuracy: 0.7940\n",
      "Epoch 26/100\n",
      "313/313 [==============================] - 25s 79ms/step - loss: 0.4312 - accuracy: 0.8019 - val_loss: 0.4392 - val_accuracy: 0.7984\n",
      "Epoch 27/100\n",
      "313/313 [==============================] - 24s 77ms/step - loss: 0.4297 - accuracy: 0.8025 - val_loss: 0.4383 - val_accuracy: 0.7978\n",
      "Epoch 28/100\n",
      "313/313 [==============================] - 24s 75ms/step - loss: 0.4290 - accuracy: 0.8045 - val_loss: 0.4390 - val_accuracy: 0.7994\n",
      "Epoch 29/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.4289 - accuracy: 0.8032 - val_loss: 0.4376 - val_accuracy: 0.7978\n",
      "Epoch 30/100\n",
      "313/313 [==============================] - 23s 75ms/step - loss: 0.4292 - accuracy: 0.8036 - val_loss: 0.4352 - val_accuracy: 0.7990\n",
      "Epoch 31/100\n",
      "313/313 [==============================] - 24s 77ms/step - loss: 0.4261 - accuracy: 0.8038 - val_loss: 0.4410 - val_accuracy: 0.7980\n",
      "Epoch 32/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4236 - accuracy: 0.8076 - val_loss: 0.4353 - val_accuracy: 0.7992\n",
      "Epoch 33/100\n",
      "313/313 [==============================] - 24s 76ms/step - loss: 0.4239 - accuracy: 0.8050 - val_loss: 0.4371 - val_accuracy: 0.8000\n"
     ]
    }
   ],
   "source": [
    "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n",
    "model3.compile(optimizer=rmsprop, loss='binary_crossentropy', \n",
    "               metrics=['accuracy'])\n",
    "\n",
    "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-2rnn-model.h5', \n",
    "                                                save_best_only=True)\n",
    "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n",
    "                                                  restore_best_weights=True)\n",
    "\n",
    "history = model3.fit(train_seq, train_target, epochs=100, batch_size=64,\n",
    "                     validation_data=(val_seq, val_target),\n",
    "                     callbacks=[checkpoint_cb, early_stopping_cb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T07:03:47.435131Z",
     "iopub.status.busy": "2022-01-08T07:03:47.434189Z",
     "iopub.status.idle": "2022-01-08T07:03:47.576747Z",
     "shell.execute_reply": "2022-01-08T07:03:47.577169Z"
    },
    "id": "1paBqczf6k2S",
    "outputId": "8dec69f9-2aa3-4b2f-f5c9-c0d2713ab4f6"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train', 'val'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Jisn0P9h6k2T"
   },
   "source": [
    "## GRU 신경망 훈련하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T07:03:47.580725Z",
     "iopub.status.busy": "2022-01-08T07:03:47.580053Z",
     "iopub.status.idle": "2022-01-08T07:03:47.802422Z",
     "shell.execute_reply": "2022-01-08T07:03:47.802844Z"
    },
    "id": "dwYF578d6k2T",
    "outputId": "72677151-359c-4d9c-9fd0-026c778f4ad5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_3 (Embedding)     (None, 100, 16)           8000      \n",
      "                                                                 \n",
      " gru (GRU)                   (None, 8)                 624       \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 1)                 9         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 8,633\n",
      "Trainable params: 8,633\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model4 = keras.Sequential()\n",
    "\n",
    "model4.add(keras.layers.Embedding(500, 16, input_length=100))\n",
    "model4.add(keras.layers.GRU(8))\n",
    "model4.add(keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "model4.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T07:03:47.818690Z",
     "iopub.status.busy": "2022-01-08T07:03:47.812276Z",
     "iopub.status.idle": "2022-01-08T07:10:35.067892Z",
     "shell.execute_reply": "2022-01-08T07:10:35.068361Z"
    },
    "id": "j2SfOZ3v6k2U",
    "outputId": "80bc83ae-1472-4c03-f372-92a6572e71b8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "313/313 [==============================] - 16s 44ms/step - loss: 0.6920 - accuracy: 0.5408 - val_loss: 0.6912 - val_accuracy: 0.5474\n",
      "Epoch 2/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.6895 - accuracy: 0.5738 - val_loss: 0.6884 - val_accuracy: 0.5764\n",
      "Epoch 3/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.6856 - accuracy: 0.5958 - val_loss: 0.6839 - val_accuracy: 0.5886\n",
      "Epoch 4/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.6794 - accuracy: 0.6119 - val_loss: 0.6771 - val_accuracy: 0.6008\n",
      "Epoch 5/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.6703 - accuracy: 0.6244 - val_loss: 0.6668 - val_accuracy: 0.6166\n",
      "Epoch 6/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.6564 - accuracy: 0.6421 - val_loss: 0.6508 - val_accuracy: 0.6352\n",
      "Epoch 7/100\n",
      "313/313 [==============================] - 13s 43ms/step - loss: 0.6350 - accuracy: 0.6601 - val_loss: 0.6252 - val_accuracy: 0.6624\n",
      "Epoch 8/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.5990 - accuracy: 0.6892 - val_loss: 0.5800 - val_accuracy: 0.7088\n",
      "Epoch 9/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.5439 - accuracy: 0.7344 - val_loss: 0.5377 - val_accuracy: 0.7320\n",
      "Epoch 10/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.5151 - accuracy: 0.7558 - val_loss: 0.5232 - val_accuracy: 0.7430\n",
      "Epoch 11/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.5008 - accuracy: 0.7656 - val_loss: 0.5108 - val_accuracy: 0.7556\n",
      "Epoch 12/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4892 - accuracy: 0.7703 - val_loss: 0.5009 - val_accuracy: 0.7636\n",
      "Epoch 13/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4788 - accuracy: 0.7796 - val_loss: 0.4948 - val_accuracy: 0.7660\n",
      "Epoch 14/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4706 - accuracy: 0.7839 - val_loss: 0.4862 - val_accuracy: 0.7758\n",
      "Epoch 15/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4628 - accuracy: 0.7889 - val_loss: 0.4807 - val_accuracy: 0.7794\n",
      "Epoch 16/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4558 - accuracy: 0.7936 - val_loss: 0.4796 - val_accuracy: 0.7706\n",
      "Epoch 17/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4508 - accuracy: 0.7966 - val_loss: 0.4728 - val_accuracy: 0.7872\n",
      "Epoch 18/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4465 - accuracy: 0.7990 - val_loss: 0.4688 - val_accuracy: 0.7870\n",
      "Epoch 19/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4426 - accuracy: 0.8016 - val_loss: 0.4663 - val_accuracy: 0.7856\n",
      "Epoch 20/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4389 - accuracy: 0.8028 - val_loss: 0.4625 - val_accuracy: 0.7884\n",
      "Epoch 21/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4362 - accuracy: 0.8045 - val_loss: 0.4628 - val_accuracy: 0.7868\n",
      "Epoch 22/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4337 - accuracy: 0.8067 - val_loss: 0.4582 - val_accuracy: 0.7900\n",
      "Epoch 23/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4311 - accuracy: 0.8084 - val_loss: 0.4573 - val_accuracy: 0.7914\n",
      "Epoch 24/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4296 - accuracy: 0.8099 - val_loss: 0.4554 - val_accuracy: 0.7918\n",
      "Epoch 25/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4274 - accuracy: 0.8116 - val_loss: 0.4569 - val_accuracy: 0.7872\n",
      "Epoch 26/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4260 - accuracy: 0.8116 - val_loss: 0.4552 - val_accuracy: 0.7908\n",
      "Epoch 27/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4249 - accuracy: 0.8109 - val_loss: 0.4595 - val_accuracy: 0.7868\n",
      "Epoch 28/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4236 - accuracy: 0.8144 - val_loss: 0.4511 - val_accuracy: 0.7908\n",
      "Epoch 29/100\n",
      "313/313 [==============================] - 13s 42ms/step - loss: 0.4227 - accuracy: 0.8134 - val_loss: 0.4534 - val_accuracy: 0.7872\n",
      "Epoch 30/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4220 - accuracy: 0.8130 - val_loss: 0.4513 - val_accuracy: 0.7920\n",
      "Epoch 31/100\n",
      "313/313 [==============================] - 13s 41ms/step - loss: 0.4205 - accuracy: 0.8154 - val_loss: 0.4520 - val_accuracy: 0.7910\n"
     ]
    }
   ],
   "source": [
    "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n",
    "model4.compile(optimizer=rmsprop, loss='binary_crossentropy', \n",
    "               metrics=['accuracy'])\n",
    "\n",
    "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-gru-model.h5', \n",
    "                                                save_best_only=True)\n",
    "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n",
    "                                                  restore_best_weights=True)\n",
    "\n",
    "history = model4.fit(train_seq, train_target, epochs=100, batch_size=64,\n",
    "                     validation_data=(val_seq, val_target),\n",
    "                     callbacks=[checkpoint_cb, early_stopping_cb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T07:10:35.080183Z",
     "iopub.status.busy": "2022-01-08T07:10:35.079530Z",
     "iopub.status.idle": "2022-01-08T07:10:35.200033Z",
     "shell.execute_reply": "2022-01-08T07:10:35.200930Z"
    },
    "id": "3yUJM4nA6k2U",
    "outputId": "cf86c235-1135-4d0f-93c2-08b7cb98ead7"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train', 'val'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ja3-116b6k2V"
   },
   "source": [
    "## 마무리"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-01-08T07:10:35.206304Z",
     "iopub.status.busy": "2022-01-08T07:10:35.205183Z",
     "iopub.status.idle": "2022-01-08T07:10:47.638916Z",
     "shell.execute_reply": "2022-01-08T07:10:47.638313Z"
    },
    "id": "-5WtFRN-6k2V",
    "outputId": "dbbf3f1d-3cd9-49cf-bf90-233ecac13134"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "782/782 [==============================] - 12s 14ms/step - loss: 0.4306 - accuracy: 0.7988\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.43062207102775574, 0.798799991607666]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_seq = pad_sequences(test_input, maxlen=100)\n",
    "\n",
    "rnn_model = keras.models.load_model('best-2rnn-model.h5')\n",
    "\n",
    "rnn_model.evaluate(test_seq, test_target)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "name": "9-3.ipynb",
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}