6.2-understanding-recurrent-neural-networks.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "G9nbXlf85Akk",
    "outputId": "2a697a04-0563-45a4-cb06-7d34f0e11d45"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'2.4.3'"
      ]
     },
     "execution_count": 1,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import keras\n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WPq89zgS5Akw"
   },
   "source": [
    "# Understanding recurrent neural networks\n",
    "\n",
    "이 노트북은 [케라스 창시자에게 배우는 딥러닝](https://tensorflow.blog/케라스-창시자에게-배우는-딥러닝/) 책의 6장 2절의 코드 예제입니다. 책에는 더 많은 내용과 그림이 있습니다. 이 노트북에는 소스 코드에 관련된 설명만 포함합니다. 이 노트북의 설명은 케라스 버전 2.2.2에 맞추어져 있습니다. 케라스 최신 버전이 릴리스되면 노트북을 다시 테스트하기 때문에 설명과 코드의 결과가 조금 다를 수 있습니다.\n",
    "\n",
    "---\n",
    "\n",
    "[...]\n",
    "\n",
    "## 케라스의 순환 층\n",
    "\n",
    "\n",
    "넘파이로 간단하게 구현한 과정이 실제 케라스의 `SimpleRNN` 층에 해당합니다:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "pG_XBpW05Akx"
   },
   "outputs": [],
   "source": [
    "from keras.layers import SimpleRNN"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "39rowLY35Aky"
   },
   "source": [
    "`SimpleRNN`이 한 가지 다른 점은 넘파이 예제처럼 하나의 시퀀스가 아니라 다른 케라스 층과 마찬가지로 시퀀스 배치를 처리한다는 것입니다. 즉, `(timesteps, input_features)` 크기가 아니라 `(batch_size, timesteps, input_features)` 크기의 입력을 받습니다.\n",
    "\n",
    "케라스에 있는 모든 순환 층과 동일하게 `SimpleRNN`은 두 가지 모드로 실행할 수 있습니다. 각 타임스텝의 출력을 모은 전체 시퀀스를 반환하거나(크기가 `(batch_size, timesteps, output_features)`인 3D 텐서), 입력 시퀀스에 대한 마지막 출력만 반환할 수 있습니다(크기가 `(batch_size, output_features)`인 2D 텐서). 이 모드는 객체를 생성할 때 `return_sequences` 매개변수로 선택할 수 있습니다. 예제를 살펴보죠:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "bGoz0doY5Aky",
    "outputId": "6f0db080-22c0-41ab-a011-6fb4e999d884"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn (SimpleRNN)       (None, 32)                2080      \n",
      "=================================================================\n",
      "Total params: 322,080\n",
      "Trainable params: 322,080\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, SimpleRNN\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7baYTpUU5Akz",
    "outputId": "855d4f84-a3ee-4fac-8a85-ebd26b97f7fb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn_1 (SimpleRNN)     (None, None, 32)          2080      \n",
      "=================================================================\n",
      "Total params: 322,080\n",
      "Trainable params: 322,080\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "nIMIVKpt5Ak0"
   },
   "source": [
    "네트워크의 표현력을 증가시키기 위해 여러 개의 순환 층을 차례대로 쌓는 것이 유용할 때가 있습니다. 이런 설정에서는 중간 층들이 전체 출력 시퀀스를 반환하도록 설정해야 합니다:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "WxopKDsU5Ak0",
    "outputId": "3844f756-3b03-4d06-f58d-c54ab45c1e1d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_2 (Embedding)      (None, None, 32)          320000    \n",
      "_________________________________________________________________\n",
      "simple_rnn_2 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_3 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_4 (SimpleRNN)     (None, None, 32)          2080      \n",
      "_________________________________________________________________\n",
      "simple_rnn_5 (SimpleRNN)     (None, 32)                2080      \n",
      "=================================================================\n",
      "Total params: 328,320\n",
      "Trainable params: 328,320\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(10000, 32))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32, return_sequences=True))\n",
    "model.add(SimpleRNN(32))  # 맨 위 층만 마지막 출력을 반환합니다.\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CQERuJcb5Ak1"
   },
   "source": [
    "이제 IMDB 영화 리뷰 분류 문제에 적용해 보죠. 먼저 데이터를 전처리합니다:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rqJFRxxD5Ak2",
    "outputId": "e4714683-2215-4e6a-e45a-3d884eb7cf95"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "데이터 로딩...\n",
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/imdb.npz\n",
      "17465344/17464789 [==============================] - 0s 0us/step\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
      "/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/datasets/imdb.py:159: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
      "  x_train, y_train = np.array(xs[:idx]), np.array(labels[:idx])\n",
      "/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/datasets/imdb.py:160: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
      "  x_test, y_test = np.array(xs[idx:]), np.array(labels[idx:])\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25000 훈련 시퀀스\n",
      "25000 테스트 시퀀스\n",
      "시퀀스 패딩 (samples x time)\n",
      "input_train 크기: (25000, 500)\n",
      "input_test 크기: (25000, 500)\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import imdb\n",
    "from keras.preprocessing import sequence\n",
    "\n",
    "max_features = 10000  # 특성으로 사용할 단어의 수\n",
    "maxlen = 500  # 사용할 텍스트의 길이(가장 빈번한 max_features 개의 단어만 사용합니다)\n",
    "batch_size = 32\n",
    "\n",
    "print('데이터 로딩...')\n",
    "(input_train, y_train), (input_test, y_test) = imdb.load_data(num_words=max_features)\n",
    "print(len(input_train), '훈련 시퀀스')\n",
    "print(len(input_test), '테스트 시퀀스')\n",
    "\n",
    "print('시퀀스 패딩 (samples x time)')\n",
    "input_train = sequence.pad_sequences(input_train, maxlen=maxlen)\n",
    "input_test = sequence.pad_sequences(input_test, maxlen=maxlen)\n",
    "print('input_train 크기:', input_train.shape)\n",
    "print('input_test 크기:', input_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WAnFNeHP5Ak2"
   },
   "source": [
    "`Embedding` 층과 `SimpleRNN` 층을 사용해 간단한 순환 네트워크를 훈련시켜 보겠습니다:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "_IzWa0e95Ak3",
    "outputId": "743c8b84-45de-440d-dbe2-fa1977ae9159"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 25s 153ms/step - loss: 0.6432 - acc: 0.6045 - val_loss: 0.4828 - val_acc: 0.7774\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 23s 145ms/step - loss: 0.3997 - acc: 0.8416 - val_loss: 0.4918 - val_acc: 0.7974\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 23s 145ms/step - loss: 0.2831 - acc: 0.8911 - val_loss: 0.4461 - val_acc: 0.7928\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 23s 146ms/step - loss: 0.2469 - acc: 0.9078 - val_loss: 0.3272 - val_acc: 0.8656\n",
      "Epoch 5/10\n",
      "157/157 [==============================] - 23s 146ms/step - loss: 0.1874 - acc: 0.9334 - val_loss: 0.3391 - val_acc: 0.8674\n",
      "Epoch 6/10\n",
      "157/157 [==============================] - 23s 146ms/step - loss: 0.1611 - acc: 0.9423 - val_loss: 0.3530 - val_acc: 0.8648\n",
      "Epoch 7/10\n",
      "157/157 [==============================] - 23s 147ms/step - loss: 0.1350 - acc: 0.9526 - val_loss: 0.4172 - val_acc: 0.8640\n",
      "Epoch 8/10\n",
      "157/157 [==============================] - 23s 146ms/step - loss: 0.1117 - acc: 0.9605 - val_loss: 0.4374 - val_acc: 0.8540\n",
      "Epoch 9/10\n",
      "157/157 [==============================] - 23s 144ms/step - loss: 0.0780 - acc: 0.9744 - val_loss: 0.4396 - val_acc: 0.8586\n",
      "Epoch 10/10\n",
      "157/157 [==============================] - 23s 143ms/step - loss: 0.0597 - acc: 0.9822 - val_loss: 0.4870 - val_acc: 0.8452\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Dense\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(max_features, 32))\n",
    "model.add(SimpleRNN(32))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])\n",
    "history = model.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "So6zp1-c5Ak3"
   },
   "source": [
    "이제 훈련과 검증의 손실과 정확도를 그래프로 그립니다:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "XF2c0ILA5Ak4"
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 545
    },
    "id": "EZqCsGtH5Ak4",
    "outputId": "e3059306-f67a-4cdc-ca21-a41bdf00be28"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_u7Fkwv95Ak5"
   },
   "source": [
    "3장에서 이 데이터셋을 사용한 첫 번째 모델에서 얻은 테스트 정확도는 87%였습니다. 안타깝지만 간단한 순환 네트워크는 이 기준 모델보다 성능이 높지 않습니다(85% 정도의 검증 정확도를 얻었습니다). 이런 원인은 전체 시퀀스가 아니라 처음 500개의 단어만 입력에 사용했기 때문입니다. 이 RNN은 기준 모델보다 얻은 정보가 적습니다. 다른 이유는 `SimpleRNN`이 텍스트와 같이 긴 시퀀스를 처리하는데 적합하지 않기 때문입니다. 더 잘 작동하는 다른 순환 층이 있습니다. 조금 더 고급 순환 층을 살펴보죠."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "iwIn16Ob5Ak5"
   },
   "source": [
    "[...]\n",
    "\n",
    "## 케라스를 사용한 LSTM 예제\n",
    "\n",
    "이제 실제적인 관심사로 이동해 보죠. LSTM 층으로 모델을 구성하고 IMDB 데이터에서 훈련해 보겠습니다(그림 6-16과 6-17 참조). 이 네트워크는 조금 전 `SimpleRNN`을 사용했던 모델과 비슷합니다. LSTM 층은 출력 차원만 지정하고 다른 (많은) 매개변수는 케라스의 기본값으로 남겨 두었습니다. 케라스는 좋은 기본값을 가지고 있어서 직접 매개변수를 튜닝하는 데 시간을 쓰지 않고도 거의 항상 어느정도 작동하는 모델을 얻을 수 있습니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "SYAq7wfy5Ak6",
    "outputId": "9d9424a6-46ee-4622-9b38-e24a27d333c5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "157/157 [==============================] - 54s 331ms/step - loss: 0.6335 - acc: 0.6474 - val_loss: 0.4211 - val_acc: 0.8188\n",
      "Epoch 2/10\n",
      "157/157 [==============================] - 55s 350ms/step - loss: 0.3103 - acc: 0.8786 - val_loss: 0.3501 - val_acc: 0.8522\n",
      "Epoch 3/10\n",
      "157/157 [==============================] - 54s 347ms/step - loss: 0.2322 - acc: 0.9124 - val_loss: 0.3130 - val_acc: 0.8678\n",
      "Epoch 4/10\n",
      "157/157 [==============================] - 58s 371ms/step - loss: 0.1964 - acc: 0.9263 - val_loss: 0.4054 - val_acc: 0.8202\n",
      "Epoch 5/10\n",
      "157/157 [==============================] - 58s 368ms/step - loss: 0.1761 - acc: 0.9338 - val_loss: 0.4619 - val_acc: 0.8286\n",
      "Epoch 6/10\n",
      "157/157 [==============================] - 58s 367ms/step - loss: 0.1583 - acc: 0.9429 - val_loss: 0.3325 - val_acc: 0.8672\n",
      "Epoch 7/10\n",
      "157/157 [==============================] - 57s 366ms/step - loss: 0.1492 - acc: 0.9484 - val_loss: 0.4155 - val_acc: 0.8356\n",
      "Epoch 8/10\n",
      "157/157 [==============================] - 58s 372ms/step - loss: 0.1292 - acc: 0.9561 - val_loss: 0.4903 - val_acc: 0.8418\n",
      "Epoch 9/10\n",
      "157/157 [==============================] - 58s 368ms/step - loss: 0.1217 - acc: 0.9601 - val_loss: 0.4022 - val_acc: 0.8760\n",
      "Epoch 10/10\n",
      "157/157 [==============================] - 58s 372ms/step - loss: 0.1097 - acc: 0.9621 - val_loss: 0.3629 - val_acc: 0.8664\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import LSTM\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(max_features, 32))\n",
    "model.add(LSTM(32))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(input_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=128,\n",
    "                    validation_split=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 545
    },
    "id": "7Vi2ksDY5Ak6",
    "outputId": "1d429091-834b-4a3a-9f91-f805720af50b"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "6.2-understanding-recurrent-neural-networks.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}