9-2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "display_name": "TensorFlow 2.3 on Python 3.6 (CUDA 10.1)",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.9"
    },
    "colab": {
      "name": "9-2.ipynb",
      "provenance": []
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MusV5YYT6kQF"
      },
      "source": [
        "# 순환 신경망으로 IMDB 리뷰 분류하기"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9uYfRb5-6kQQ"
      },
      "source": [
        "<table align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/rickiepark/hg-mldl/blob/master/9-2.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": "S1A-DDIQ6kQQ"
      },
      "source": [
        "## IMDB 리뷰 데이터셋"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:41.979392Z",
          "iopub.status.busy": "2020-10-12T01:18:41.978299Z",
          "iopub.status.idle": "2020-10-12T01:18:52.473300Z",
          "shell.execute_reply": "2020-10-12T01:18:52.471703Z"
        },
        "id": "sZoZm1ZO6kQR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c340dd55-629f-4615-db76-9c06bc46e2c2"
      },
      "source": [
        "from tensorflow.keras.datasets import imdb\n",
        "\n",
        "(train_input, train_target), (test_input, test_target) = imdb.load_data(\n",
        "    num_words=500)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/imdb.npz\n",
            "17465344/17464789 [==============================] - 0s 0us/step\n",
            "17473536/17464789 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.481400Z",
          "iopub.status.busy": "2020-10-12T01:18:52.479645Z",
          "iopub.status.idle": "2020-10-12T01:18:52.485658Z",
          "shell.execute_reply": "2020-10-12T01:18:52.486465Z"
        },
        "id": "HTFjz6k06kQR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1d4929d7-0a93-4bb8-cf03-682ada3d6ac9"
      },
      "source": [
        "print(train_input.shape, test_input.shape)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(25000,) (25000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.493608Z",
          "iopub.status.busy": "2020-10-12T01:18:52.492625Z",
          "iopub.status.idle": "2020-10-12T01:18:52.496804Z",
          "shell.execute_reply": "2020-10-12T01:18:52.497722Z"
        },
        "id": "4jMLURzw6kQS",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ffb635f3-a3bd-42ed-b6da-fecaa0ede408"
      },
      "source": [
        "print(len(train_input[0]))"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "218\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.504933Z",
          "iopub.status.busy": "2020-10-12T01:18:52.503811Z",
          "iopub.status.idle": "2020-10-12T01:18:52.508252Z",
          "shell.execute_reply": "2020-10-12T01:18:52.509172Z"
        },
        "id": "xdO834PL6kQT",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "28ba2580-b6b8-49b0-e07e-4df8026a5626"
      },
      "source": [
        "print(len(train_input[1]))"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "189\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.516102Z",
          "iopub.status.busy": "2020-10-12T01:18:52.515014Z",
          "iopub.status.idle": "2020-10-12T01:18:52.520096Z",
          "shell.execute_reply": "2020-10-12T01:18:52.520972Z"
        },
        "id": "QNuzGQ2c6kQT",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "88f7dd8c-188e-4ad9-edb2-f4b59c498d40"
      },
      "source": [
        "print(train_input[0])"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1, 14, 22, 16, 43, 2, 2, 2, 2, 65, 458, 2, 66, 2, 4, 173, 36, 256, 5, 25, 100, 43, 2, 112, 50, 2, 2, 9, 35, 480, 284, 5, 150, 4, 172, 112, 167, 2, 336, 385, 39, 4, 172, 2, 2, 17, 2, 38, 13, 447, 4, 192, 50, 16, 6, 147, 2, 19, 14, 22, 4, 2, 2, 469, 4, 22, 71, 87, 12, 16, 43, 2, 38, 76, 15, 13, 2, 4, 22, 17, 2, 17, 12, 16, 2, 18, 2, 5, 62, 386, 12, 8, 316, 8, 106, 5, 4, 2, 2, 16, 480, 66, 2, 33, 4, 130, 12, 16, 38, 2, 5, 25, 124, 51, 36, 135, 48, 25, 2, 33, 6, 22, 12, 215, 28, 77, 52, 5, 14, 407, 16, 82, 2, 8, 4, 107, 117, 2, 15, 256, 4, 2, 7, 2, 5, 2, 36, 71, 43, 2, 476, 26, 400, 317, 46, 7, 4, 2, 2, 13, 104, 88, 4, 381, 15, 297, 98, 32, 2, 56, 26, 141, 6, 194, 2, 18, 4, 226, 22, 21, 134, 476, 26, 480, 5, 144, 30, 2, 18, 51, 36, 28, 224, 92, 25, 104, 4, 226, 65, 16, 38, 2, 88, 12, 16, 283, 5, 16, 2, 113, 103, 32, 15, 16, 2, 19, 178, 32]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.528598Z",
          "iopub.status.busy": "2020-10-12T01:18:52.527510Z",
          "iopub.status.idle": "2020-10-12T01:18:52.532011Z",
          "shell.execute_reply": "2020-10-12T01:18:52.532936Z"
        },
        "id": "In8Qzltz6kQT",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "78b5ddc6-d634-4755-9435-25cc8f2c9462"
      },
      "source": [
        "print(train_target[:20])"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1 0 0 1 0 0 1 0 1 0 1 0 0 0 0 0 1 1 0 1]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.541031Z",
          "iopub.status.busy": "2020-10-12T01:18:52.539873Z",
          "iopub.status.idle": "2020-10-12T01:18:52.812559Z",
          "shell.execute_reply": "2020-10-12T01:18:52.810985Z"
        },
        "id": "0-zjNgBw6kQU"
      },
      "source": [
        "from sklearn.model_selection import train_test_split\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)"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.833360Z",
          "iopub.status.busy": "2020-10-12T01:18:52.831841Z",
          "iopub.status.idle": "2020-10-12T01:18:52.834920Z",
          "shell.execute_reply": "2020-10-12T01:18:52.836117Z"
        },
        "id": "sRVvgntx6kQU"
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "lengths = np.array([len(x) for x in train_input])"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.843759Z",
          "iopub.status.busy": "2020-10-12T01:18:52.842297Z",
          "iopub.status.idle": "2020-10-12T01:18:52.847506Z",
          "shell.execute_reply": "2020-10-12T01:18:52.848361Z"
        },
        "id": "p76m_61w6kQU",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "04b533f6-45ef-4514-d87d-b484a89afde6"
      },
      "source": [
        "print(np.mean(lengths), np.median(lengths))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "239.00925 178.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:52.874891Z",
          "iopub.status.busy": "2020-10-12T01:18:52.873857Z",
          "iopub.status.idle": "2020-10-12T01:18:53.042560Z",
          "shell.execute_reply": "2020-10-12T01:18:53.040898Z"
        },
        "id": "TVDrsRYF6kQV",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "2610932c-ac31-480b-8a6c-fd5ac2793a34"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.hist(lengths)\n",
        "plt.xlabel('length')\n",
        "plt.ylabel('frequency')\n",
        "plt.show()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.056410Z",
          "iopub.status.busy": "2020-10-12T01:18:53.054676Z",
          "iopub.status.idle": "2020-10-12T01:18:53.509881Z",
          "shell.execute_reply": "2020-10-12T01:18:53.507938Z"
        },
        "id": "fbLSFToP6kQV"
      },
      "source": [
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "\n",
        "train_seq = pad_sequences(train_input, maxlen=100)"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.516113Z",
          "iopub.status.busy": "2020-10-12T01:18:53.514858Z",
          "iopub.status.idle": "2020-10-12T01:18:53.519062Z",
          "shell.execute_reply": "2020-10-12T01:18:53.520171Z"
        },
        "id": "3c86x0Yj6kQV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "52d4ff3a-9aa5-4ebd-e086-d9b2d490a29f"
      },
      "source": [
        "print(train_seq.shape)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(20000, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.529088Z",
          "iopub.status.busy": "2020-10-12T01:18:53.527775Z",
          "iopub.status.idle": "2020-10-12T01:18:53.532885Z",
          "shell.execute_reply": "2020-10-12T01:18:53.533957Z"
        },
        "id": "JuP0QJaU6kQW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "95e6521e-b200-4733-f5e8-65320ff2a81f"
      },
      "source": [
        "print(train_seq[0])"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[ 10   4  20   9   2 364 352   5  45   6   2   2  33 269   8   2 142   2\n",
            "   5   2  17  73  17 204   5   2  19  55   2   2  92  66 104  14  20  93\n",
            "  76   2 151  33   4  58  12 188   2 151  12 215  69 224 142  73 237   6\n",
            "   2   7   2   2 188   2 103  14  31  10  10 451   7   2   5   2  80  91\n",
            "   2  30   2  34  14  20 151  50  26 131  49   2  84  46  50  37  80  79\n",
            "   6   2  46   7  14  20  10  10 470 158]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.541335Z",
          "iopub.status.busy": "2020-10-12T01:18:53.540161Z",
          "iopub.status.idle": "2020-10-12T01:18:53.544735Z",
          "shell.execute_reply": "2020-10-12T01:18:53.545765Z"
        },
        "id": "6t8kaDji6kQW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a9d5ba15-838d-4cac-c23c-18c2fa4b709f"
      },
      "source": [
        "print(train_input[0][-10:])"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[6, 2, 46, 7, 14, 20, 10, 10, 470, 158]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.553993Z",
          "iopub.status.busy": "2020-10-12T01:18:53.552689Z",
          "iopub.status.idle": "2020-10-12T01:18:53.557009Z",
          "shell.execute_reply": "2020-10-12T01:18:53.558118Z"
        },
        "scrolled": true,
        "id": "Emmoztwa6kQX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d72b6139-e876-41a1-b331-313d3199548a"
      },
      "source": [
        "print(train_seq[5])"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[  0   0   0   0   1   2 195  19  49   2   2 190   4   2 352   2 183  10\n",
            "  10  13  82  79   4   2  36  71 269   8   2  25  19  49   7   4   2   2\n",
            "   2   2   2  10  10  48  25  40   2  11   2   2  40   2   2   5   4   2\n",
            "   2  95  14 238  56 129   2  10  10  21   2  94 364 352   2   2  11 190\n",
            "  24 484   2   7  94 205 405  10  10  87   2  34  49   2   7   2   2   2\n",
            "   2   2 290   2  46  48  64  18   4   2]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.568279Z",
          "iopub.status.busy": "2020-10-12T01:18:53.566949Z",
          "iopub.status.idle": "2020-10-12T01:18:53.688571Z",
          "shell.execute_reply": "2020-10-12T01:18:53.687412Z"
        },
        "id": "ZeFmsjOd6kQX"
      },
      "source": [
        "val_seq = pad_sequences(val_input, maxlen=100)"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9xab0nu_6kQX"
      },
      "source": [
        "## 순환 신경망 만들기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:53.702727Z",
          "iopub.status.busy": "2020-10-12T01:18:53.694374Z",
          "iopub.status.idle": "2020-10-12T01:18:55.997285Z",
          "shell.execute_reply": "2020-10-12T01:18:55.995617Z"
        },
        "id": "45BtHGKX6kQX"
      },
      "source": [
        "from tensorflow import keras\n",
        "\n",
        "model = keras.Sequential()\n",
        "\n",
        "model.add(keras.layers.SimpleRNN(8, input_shape=(100, 500)))\n",
        "model.add(keras.layers.Dense(1, activation='sigmoid'))"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:56.004523Z",
          "iopub.status.busy": "2020-10-12T01:18:56.002786Z",
          "iopub.status.idle": "2020-10-12T01:18:57.766487Z",
          "shell.execute_reply": "2020-10-12T01:18:57.764928Z"
        },
        "id": "oXIWyOuw6kQY"
      },
      "source": [
        "train_oh = keras.utils.to_categorical(train_seq)"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:57.774226Z",
          "iopub.status.busy": "2020-10-12T01:18:57.772482Z",
          "iopub.status.idle": "2020-10-12T01:18:57.777731Z",
          "shell.execute_reply": "2020-10-12T01:18:57.778538Z"
        },
        "id": "VD94WiRT6kQY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d2074cc9-cdb3-4c73-e9ca-1318fdd34034"
      },
      "source": [
        "print(train_oh.shape)"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(20000, 100, 500)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:57.786369Z",
          "iopub.status.busy": "2020-10-12T01:18:57.785416Z",
          "iopub.status.idle": "2020-10-12T01:18:57.790217Z",
          "shell.execute_reply": "2020-10-12T01:18:57.791183Z"
        },
        "id": "VIVSHwJg6kQY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "972653fa-6acd-405d-d1ee-2422708f9863"
      },
      "source": [
        "print(train_oh[0][0][:12])"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:57.798946Z",
          "iopub.status.busy": "2020-10-12T01:18:57.797788Z",
          "iopub.status.idle": "2020-10-12T01:18:57.802480Z",
          "shell.execute_reply": "2020-10-12T01:18:57.803390Z"
        },
        "id": "CP3ivii46kQZ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6a669f1a-41fe-486a-e0e1-bd812d2b7aeb"
      },
      "source": [
        "print(np.sum(train_oh[0][0]))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:57.810383Z",
          "iopub.status.busy": "2020-10-12T01:18:57.809293Z",
          "iopub.status.idle": "2020-10-12T01:18:58.247989Z",
          "shell.execute_reply": "2020-10-12T01:18:58.248931Z"
        },
        "id": "vgijH6m26kQZ"
      },
      "source": [
        "val_oh = keras.utils.to_categorical(val_seq)"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:58.256367Z",
          "iopub.status.busy": "2020-10-12T01:18:58.255252Z",
          "iopub.status.idle": "2020-10-12T01:18:58.261986Z",
          "shell.execute_reply": "2020-10-12T01:18:58.262846Z"
        },
        "id": "0V0Svpuf6kQZ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "34f0dd62-57da-4907-8032-9bb2655a986e"
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "simple_rnn (SimpleRNN)       (None, 8)                 4072      \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 1)                 9         \n",
            "=================================================================\n",
            "Total params: 4,081\n",
            "Trainable params: 4,081\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S22Hc-7L6kQZ"
      },
      "source": [
        "## 순환 신경망 훈련하기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:18:58.286672Z",
          "iopub.status.busy": "2020-10-12T01:18:58.285571Z",
          "iopub.status.idle": "2020-10-12T01:30:36.157752Z",
          "shell.execute_reply": "2020-10-12T01:30:36.159280Z"
        },
        "id": "YcwM299t6kQZ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "257518bf-52ab-49d8-aa91-3c2e606ffd74"
      },
      "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-simplernn-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_oh, train_target, epochs=100, batch_size=64,\n",
        "                    validation_data=(val_oh, val_target),\n",
        "                    callbacks=[checkpoint_cb, early_stopping_cb])"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "313/313 [==============================] - 38s 110ms/step - loss: 0.6964 - accuracy: 0.4965 - val_loss: 0.6961 - val_accuracy: 0.4894\n",
            "Epoch 2/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.6938 - accuracy: 0.5041 - val_loss: 0.6945 - val_accuracy: 0.4946\n",
            "Epoch 3/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.6923 - accuracy: 0.5143 - val_loss: 0.6933 - val_accuracy: 0.5078\n",
            "Epoch 4/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6911 - accuracy: 0.5243 - val_loss: 0.6926 - val_accuracy: 0.5126\n",
            "Epoch 5/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.6900 - accuracy: 0.5321 - val_loss: 0.6917 - val_accuracy: 0.5252\n",
            "Epoch 6/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6889 - accuracy: 0.5411 - val_loss: 0.6909 - val_accuracy: 0.5362\n",
            "Epoch 7/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6876 - accuracy: 0.5484 - val_loss: 0.6900 - val_accuracy: 0.5398\n",
            "Epoch 8/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6862 - accuracy: 0.5567 - val_loss: 0.6891 - val_accuracy: 0.5430\n",
            "Epoch 9/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6846 - accuracy: 0.5613 - val_loss: 0.6879 - val_accuracy: 0.5450\n",
            "Epoch 10/100\n",
            "313/313 [==============================] - 30s 97ms/step - loss: 0.6824 - accuracy: 0.5688 - val_loss: 0.6859 - val_accuracy: 0.5484\n",
            "Epoch 11/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.6769 - accuracy: 0.5925 - val_loss: 0.6737 - val_accuracy: 0.6056\n",
            "Epoch 12/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.6597 - accuracy: 0.6319 - val_loss: 0.6562 - val_accuracy: 0.6336\n",
            "Epoch 13/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.6371 - accuracy: 0.6592 - val_loss: 0.6310 - val_accuracy: 0.6638\n",
            "Epoch 14/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.6102 - accuracy: 0.6849 - val_loss: 0.6020 - val_accuracy: 0.6868\n",
            "Epoch 15/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.5838 - accuracy: 0.7116 - val_loss: 0.5793 - val_accuracy: 0.7150\n",
            "Epoch 16/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.5634 - accuracy: 0.7286 - val_loss: 0.5618 - val_accuracy: 0.7316\n",
            "Epoch 17/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.5454 - accuracy: 0.7426 - val_loss: 0.5533 - val_accuracy: 0.7312\n",
            "Epoch 18/100\n",
            "313/313 [==============================] - 34s 107ms/step - loss: 0.5293 - accuracy: 0.7518 - val_loss: 0.5317 - val_accuracy: 0.7460\n",
            "Epoch 19/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.5157 - accuracy: 0.7616 - val_loss: 0.5197 - val_accuracy: 0.7540\n",
            "Epoch 20/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.5032 - accuracy: 0.7692 - val_loss: 0.5100 - val_accuracy: 0.7604\n",
            "Epoch 21/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.4925 - accuracy: 0.7742 - val_loss: 0.5008 - val_accuracy: 0.7660\n",
            "Epoch 22/100\n",
            "313/313 [==============================] - 33s 106ms/step - loss: 0.4823 - accuracy: 0.7801 - val_loss: 0.4973 - val_accuracy: 0.7602\n",
            "Epoch 23/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4744 - accuracy: 0.7846 - val_loss: 0.4908 - val_accuracy: 0.7692\n",
            "Epoch 24/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4667 - accuracy: 0.7888 - val_loss: 0.4813 - val_accuracy: 0.7742\n",
            "Epoch 25/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4601 - accuracy: 0.7923 - val_loss: 0.4752 - val_accuracy: 0.7778\n",
            "Epoch 26/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4540 - accuracy: 0.7959 - val_loss: 0.4733 - val_accuracy: 0.7774\n",
            "Epoch 27/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4491 - accuracy: 0.7980 - val_loss: 0.4666 - val_accuracy: 0.7874\n",
            "Epoch 28/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4439 - accuracy: 0.7995 - val_loss: 0.4650 - val_accuracy: 0.7808\n",
            "Epoch 29/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4391 - accuracy: 0.8036 - val_loss: 0.4639 - val_accuracy: 0.7866\n",
            "Epoch 30/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.4353 - accuracy: 0.8054 - val_loss: 0.4588 - val_accuracy: 0.7876\n",
            "Epoch 31/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.4319 - accuracy: 0.8081 - val_loss: 0.4585 - val_accuracy: 0.7844\n",
            "Epoch 32/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4287 - accuracy: 0.8080 - val_loss: 0.4574 - val_accuracy: 0.7812\n",
            "Epoch 33/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.4259 - accuracy: 0.8104 - val_loss: 0.4551 - val_accuracy: 0.7912\n",
            "Epoch 34/100\n",
            "313/313 [==============================] - 33s 104ms/step - loss: 0.4237 - accuracy: 0.8125 - val_loss: 0.4528 - val_accuracy: 0.7880\n",
            "Epoch 35/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4211 - accuracy: 0.8131 - val_loss: 0.4512 - val_accuracy: 0.7918\n",
            "Epoch 36/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.4185 - accuracy: 0.8159 - val_loss: 0.4527 - val_accuracy: 0.7878\n",
            "Epoch 37/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4174 - accuracy: 0.8153 - val_loss: 0.4497 - val_accuracy: 0.7940\n",
            "Epoch 38/100\n",
            "313/313 [==============================] - 32s 103ms/step - loss: 0.4152 - accuracy: 0.8150 - val_loss: 0.4591 - val_accuracy: 0.7902\n",
            "Epoch 39/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4138 - accuracy: 0.8177 - val_loss: 0.4494 - val_accuracy: 0.7928\n",
            "Epoch 40/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.4118 - accuracy: 0.8170 - val_loss: 0.4478 - val_accuracy: 0.7940\n",
            "Epoch 41/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4102 - accuracy: 0.8171 - val_loss: 0.4475 - val_accuracy: 0.7956\n",
            "Epoch 42/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4090 - accuracy: 0.8189 - val_loss: 0.4471 - val_accuracy: 0.7976\n",
            "Epoch 43/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4079 - accuracy: 0.8199 - val_loss: 0.4466 - val_accuracy: 0.7934\n",
            "Epoch 44/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4074 - accuracy: 0.8205 - val_loss: 0.4455 - val_accuracy: 0.7938\n",
            "Epoch 45/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4057 - accuracy: 0.8207 - val_loss: 0.4461 - val_accuracy: 0.7966\n",
            "Epoch 46/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4050 - accuracy: 0.8206 - val_loss: 0.4522 - val_accuracy: 0.7962\n",
            "Epoch 47/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.4038 - accuracy: 0.8239 - val_loss: 0.4447 - val_accuracy: 0.7936\n",
            "Epoch 48/100\n",
            "313/313 [==============================] - 32s 102ms/step - loss: 0.4027 - accuracy: 0.8242 - val_loss: 0.4491 - val_accuracy: 0.7964\n",
            "Epoch 49/100\n",
            "313/313 [==============================] - 32s 101ms/step - loss: 0.4019 - accuracy: 0.8230 - val_loss: 0.4496 - val_accuracy: 0.7938\n",
            "Epoch 50/100\n",
            "313/313 [==============================] - 31s 100ms/step - loss: 0.4009 - accuracy: 0.8242 - val_loss: 0.4461 - val_accuracy: 0.7968\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:30:36.190161Z",
          "iopub.status.busy": "2020-10-12T01:30:36.175046Z",
          "iopub.status.idle": "2020-10-12T01:30:36.782577Z",
          "shell.execute_reply": "2020-10-12T01:30:36.784046Z"
        },
        "id": "YS_5tIG26kQa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "7affeec2-15fb-4217-daa5-2c8ef2f8a74b"
      },
      "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()"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K_U_lVie6kQa"
      },
      "source": [
        "## 단어 임베딩을 사용하기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:30:36.795147Z",
          "iopub.status.busy": "2020-10-12T01:30:36.793959Z",
          "iopub.status.idle": "2020-10-12T01:30:37.414822Z",
          "shell.execute_reply": "2020-10-12T01:30:37.417021Z"
        },
        "id": "Ofl-Ifev6kQa",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5446f8dd-845c-4fea-eb3a-a8887f010da0"
      },
      "source": [
        "model2 = keras.Sequential()\n",
        "\n",
        "model2.add(keras.layers.Embedding(500, 16, input_length=100))\n",
        "model2.add(keras.layers.SimpleRNN(8))\n",
        "model2.add(keras.layers.Dense(1, activation='sigmoid'))\n",
        "\n",
        "model2.summary()"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_1 (Embedding)      (None, 100, 16)           8000      \n",
            "_________________________________________________________________\n",
            "simple_rnn_2 (SimpleRNN)     (None, 8)                 200       \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 1)                 9         \n",
            "=================================================================\n",
            "Total params: 8,209\n",
            "Trainable params: 8,209\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:30:37.439013Z",
          "iopub.status.busy": "2020-10-12T01:30:37.437655Z",
          "iopub.status.idle": "2020-10-12T01:40:38.584921Z",
          "shell.execute_reply": "2020-10-12T01:40:38.586340Z"
        },
        "id": "78TUrJAs6kQa",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ee782c94-cf81-4add-e0a9-5290b91bd2bf"
      },
      "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-embedding-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])"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "313/313 [==============================] - 45s 138ms/step - loss: 0.6885 - accuracy: 0.5390 - val_loss: 0.6777 - val_accuracy: 0.5888\n",
            "Epoch 2/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.6352 - accuracy: 0.6608 - val_loss: 0.6000 - val_accuracy: 0.7100\n",
            "Epoch 3/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.5780 - accuracy: 0.7263 - val_loss: 0.5707 - val_accuracy: 0.7300\n",
            "Epoch 4/100\n",
            "313/313 [==============================] - 42s 135ms/step - loss: 0.5466 - accuracy: 0.7505 - val_loss: 0.5402 - val_accuracy: 0.7486\n",
            "Epoch 5/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.5244 - accuracy: 0.7669 - val_loss: 0.5316 - val_accuracy: 0.7550\n",
            "Epoch 6/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.5079 - accuracy: 0.7767 - val_loss: 0.5142 - val_accuracy: 0.7678\n",
            "Epoch 7/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.4963 - accuracy: 0.7835 - val_loss: 0.5149 - val_accuracy: 0.7598\n",
            "Epoch 8/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.4856 - accuracy: 0.7880 - val_loss: 0.4977 - val_accuracy: 0.7740\n",
            "Epoch 9/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4761 - accuracy: 0.7926 - val_loss: 0.4901 - val_accuracy: 0.7790\n",
            "Epoch 10/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.4694 - accuracy: 0.7940 - val_loss: 0.4869 - val_accuracy: 0.7766\n",
            "Epoch 11/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4625 - accuracy: 0.7993 - val_loss: 0.4812 - val_accuracy: 0.7810\n",
            "Epoch 12/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.4574 - accuracy: 0.7982 - val_loss: 0.4973 - val_accuracy: 0.7610\n",
            "Epoch 13/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4531 - accuracy: 0.8032 - val_loss: 0.4786 - val_accuracy: 0.7802\n",
            "Epoch 14/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4483 - accuracy: 0.8044 - val_loss: 0.4758 - val_accuracy: 0.7832\n",
            "Epoch 15/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4447 - accuracy: 0.8062 - val_loss: 0.4707 - val_accuracy: 0.7882\n",
            "Epoch 16/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4408 - accuracy: 0.8094 - val_loss: 0.4673 - val_accuracy: 0.7886\n",
            "Epoch 17/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.4378 - accuracy: 0.8080 - val_loss: 0.4704 - val_accuracy: 0.7810\n",
            "Epoch 18/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.4341 - accuracy: 0.8105 - val_loss: 0.4627 - val_accuracy: 0.7898\n",
            "Epoch 19/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.4319 - accuracy: 0.8105 - val_loss: 0.4617 - val_accuracy: 0.7878\n",
            "Epoch 20/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.4293 - accuracy: 0.8128 - val_loss: 0.4628 - val_accuracy: 0.7868\n",
            "Epoch 21/100\n",
            "313/313 [==============================] - 43s 136ms/step - loss: 0.4274 - accuracy: 0.8120 - val_loss: 0.4607 - val_accuracy: 0.7854\n",
            "Epoch 22/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4247 - accuracy: 0.8148 - val_loss: 0.4628 - val_accuracy: 0.7872\n",
            "Epoch 23/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4223 - accuracy: 0.8155 - val_loss: 0.4581 - val_accuracy: 0.7882\n",
            "Epoch 24/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4209 - accuracy: 0.8152 - val_loss: 0.4624 - val_accuracy: 0.7856\n",
            "Epoch 25/100\n",
            "313/313 [==============================] - 43s 138ms/step - loss: 0.4179 - accuracy: 0.8168 - val_loss: 0.4590 - val_accuracy: 0.7862\n",
            "Epoch 26/100\n",
            "313/313 [==============================] - 43s 137ms/step - loss: 0.4156 - accuracy: 0.8181 - val_loss: 0.4585 - val_accuracy: 0.7888\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2020-10-12T01:40:38.604725Z",
          "iopub.status.busy": "2020-10-12T01:40:38.602954Z",
          "iopub.status.idle": "2020-10-12T01:40:39.220316Z",
          "shell.execute_reply": "2020-10-12T01:40:39.221611Z"
        },
        "id": "brUvKwLt6kQb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "c9907bb0-000c-4c7f-fe2c-ce8128bac52b"
      },
      "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()"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}