material/deep_learning/treating_overfitting_with_imdb.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "treating_overfitting_with_imdb.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "fTFj8ft5dlbS"
      ]
    },
    "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": "3.8.8"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C9HmC2T4ld5B"
      },
      "source": [
        "# 오버피팅 처리. IMDB 데이터\n",
        "\n",
        "copy from https://www.tensorflow.org/tutorials/keras/overfit_and_underfit?hl=ko\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5pZ8A2liqvgk"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-d3335KZKWke"
      },
      "source": [
        "# 데이터 준비"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aRl8BwQhKVSy"
      },
      "source": [
        "NUM_WORDS = 1000"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PkZfm42fKyx5",
        "outputId": "aeca07d2-769d-4b1d-9238-c0836edfec32"
      },
      "source": [
        "(raw_train_x, raw_train_y), (raw_test_x, raw_test_y) = tf.keras.datasets.imdb.load_data(num_words=NUM_WORDS)\n",
        "\n",
        "print(raw_train_x.shape)\n",
        "print(raw_train_y.shape)\n",
        "print(raw_test_x.shape)\n",
        "print(raw_test_y.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(25000,)\n",
            "(25000,)\n",
            "(25000,)\n",
            "(25000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CO1KFBtwK84Z",
        "outputId": "d0d36970-d900-4db9-89f2-ca171b95a241"
      },
      "source": [
        "print(raw_train_x[:5])\n",
        "print(raw_train_y[:5])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[list([1, 14, 22, 16, 43, 530, 973, 2, 2, 65, 458, 2, 66, 2, 4, 173, 36, 256, 5, 25, 100, 43, 838, 112, 50, 670, 2, 9, 35, 480, 284, 5, 150, 4, 172, 112, 167, 2, 336, 385, 39, 4, 172, 2, 2, 17, 546, 38, 13, 447, 4, 192, 50, 16, 6, 147, 2, 19, 14, 22, 4, 2, 2, 469, 4, 22, 71, 87, 12, 16, 43, 530, 38, 76, 15, 13, 2, 4, 22, 17, 515, 17, 12, 16, 626, 18, 2, 5, 62, 386, 12, 8, 316, 8, 106, 5, 4, 2, 2, 16, 480, 66, 2, 33, 4, 130, 12, 16, 38, 619, 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, 723, 36, 71, 43, 530, 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",
            " list([1, 194, 2, 194, 2, 78, 228, 5, 6, 2, 2, 2, 134, 26, 4, 715, 8, 118, 2, 14, 394, 20, 13, 119, 954, 189, 102, 5, 207, 110, 2, 21, 14, 69, 188, 8, 30, 23, 7, 4, 249, 126, 93, 4, 114, 9, 2, 2, 5, 647, 4, 116, 9, 35, 2, 4, 229, 9, 340, 2, 4, 118, 9, 4, 130, 2, 19, 4, 2, 5, 89, 29, 952, 46, 37, 4, 455, 9, 45, 43, 38, 2, 2, 398, 4, 2, 26, 2, 5, 163, 11, 2, 2, 4, 2, 9, 194, 775, 7, 2, 2, 349, 2, 148, 605, 2, 2, 15, 123, 125, 68, 2, 2, 15, 349, 165, 2, 98, 5, 4, 228, 9, 43, 2, 2, 15, 299, 120, 5, 120, 174, 11, 220, 175, 136, 50, 9, 2, 228, 2, 5, 2, 656, 245, 2, 5, 4, 2, 131, 152, 491, 18, 2, 32, 2, 2, 14, 9, 6, 371, 78, 22, 625, 64, 2, 9, 8, 168, 145, 23, 4, 2, 15, 16, 4, 2, 5, 28, 6, 52, 154, 462, 33, 89, 78, 285, 16, 145, 95])\n",
            " list([1, 14, 47, 8, 30, 31, 7, 4, 249, 108, 7, 4, 2, 54, 61, 369, 13, 71, 149, 14, 22, 112, 4, 2, 311, 12, 16, 2, 33, 75, 43, 2, 296, 4, 86, 320, 35, 534, 19, 263, 2, 2, 4, 2, 33, 89, 78, 12, 66, 16, 4, 360, 7, 4, 58, 316, 334, 11, 4, 2, 43, 645, 662, 8, 257, 85, 2, 42, 2, 2, 83, 68, 2, 15, 36, 165, 2, 278, 36, 69, 2, 780, 8, 106, 14, 2, 2, 18, 6, 22, 12, 215, 28, 610, 40, 6, 87, 326, 23, 2, 21, 23, 22, 12, 272, 40, 57, 31, 11, 4, 22, 47, 6, 2, 51, 9, 170, 23, 595, 116, 595, 2, 13, 191, 79, 638, 89, 2, 14, 9, 8, 106, 607, 624, 35, 534, 6, 227, 7, 129, 113])\n",
            " list([1, 4, 2, 2, 33, 2, 4, 2, 432, 111, 153, 103, 4, 2, 13, 70, 131, 67, 11, 61, 2, 744, 35, 2, 761, 61, 2, 452, 2, 4, 985, 7, 2, 59, 166, 4, 105, 216, 2, 41, 2, 9, 15, 7, 35, 744, 2, 31, 8, 4, 687, 23, 4, 2, 2, 6, 2, 42, 38, 39, 121, 59, 456, 10, 10, 7, 265, 12, 575, 111, 153, 159, 59, 16, 2, 21, 25, 586, 482, 39, 4, 96, 59, 716, 12, 4, 172, 65, 9, 579, 11, 2, 4, 2, 5, 2, 7, 2, 17, 13, 2, 12, 19, 6, 464, 31, 314, 11, 2, 6, 719, 605, 11, 8, 202, 27, 310, 4, 2, 2, 8, 2, 58, 10, 10, 537, 2, 180, 40, 14, 413, 173, 7, 263, 112, 37, 152, 377, 4, 537, 263, 846, 579, 178, 54, 75, 71, 476, 36, 413, 263, 2, 182, 5, 17, 75, 2, 922, 36, 279, 131, 2, 17, 2, 42, 17, 35, 921, 2, 192, 5, 2, 2, 19, 2, 217, 2, 2, 537, 2, 2, 5, 736, 10, 10, 61, 403, 9, 2, 40, 61, 2, 5, 27, 2, 159, 90, 263, 2, 2, 309, 8, 178, 5, 82, 2, 4, 65, 15, 2, 145, 143, 2, 12, 2, 537, 746, 537, 537, 15, 2, 4, 2, 594, 7, 2, 94, 2, 2, 2, 11, 2, 4, 538, 7, 2, 246, 2, 9, 2, 11, 635, 14, 9, 51, 408, 12, 94, 318, 2, 12, 47, 6, 2, 936, 5, 2, 2, 19, 49, 7, 4, 2, 2, 2, 25, 80, 126, 842, 10, 10, 2, 2, 2, 27, 2, 11, 2, 2, 159, 27, 341, 29, 2, 19, 2, 173, 7, 90, 2, 8, 30, 11, 4, 2, 86, 2, 8, 2, 46, 11, 2, 21, 29, 9, 2, 23, 4, 2, 2, 793, 6, 2, 2, 2, 10, 10, 246, 50, 9, 6, 2, 2, 746, 90, 29, 2, 8, 124, 4, 882, 4, 882, 496, 27, 2, 2, 537, 121, 127, 2, 130, 5, 29, 494, 8, 124, 4, 882, 496, 4, 341, 7, 27, 846, 10, 10, 29, 9, 2, 8, 97, 6, 236, 2, 2, 8, 4, 2, 7, 31, 7, 2, 91, 2, 2, 70, 4, 882, 30, 579, 42, 9, 12, 32, 11, 537, 10, 10, 11, 14, 65, 44, 537, 75, 2, 2, 2, 2, 2, 4, 2, 7, 154, 5, 4, 518, 53, 2, 2, 7, 2, 882, 11, 399, 38, 75, 257, 2, 19, 2, 17, 29, 456, 4, 65, 7, 27, 205, 113, 10, 10, 2, 4, 2, 2, 9, 242, 4, 91, 2, 2, 5, 2, 307, 22, 7, 2, 126, 93, 40, 2, 13, 188, 2, 2, 19, 4, 2, 7, 2, 537, 23, 53, 537, 21, 82, 40, 2, 13, 2, 14, 280, 13, 219, 4, 2, 431, 758, 859, 4, 953, 2, 2, 7, 2, 5, 94, 40, 25, 238, 60, 2, 4, 2, 804, 2, 7, 4, 2, 132, 8, 67, 6, 22, 15, 9, 283, 8, 2, 14, 31, 9, 242, 955, 48, 25, 279, 2, 23, 12, 2, 195, 25, 238, 60, 796, 2, 4, 671, 7, 2, 5, 4, 559, 154, 888, 7, 726, 50, 26, 49, 2, 15, 566, 30, 579, 21, 64, 2])\n",
            " list([1, 249, 2, 7, 61, 113, 10, 10, 13, 2, 14, 20, 56, 33, 2, 18, 457, 88, 13, 2, 2, 45, 2, 13, 70, 79, 49, 706, 919, 13, 16, 355, 340, 355, 2, 96, 143, 4, 22, 32, 289, 7, 61, 369, 71, 2, 5, 13, 16, 131, 2, 249, 114, 249, 229, 249, 20, 13, 28, 126, 110, 13, 473, 8, 569, 61, 419, 56, 429, 6, 2, 18, 35, 534, 95, 474, 570, 5, 25, 124, 138, 88, 12, 421, 2, 52, 725, 2, 61, 419, 11, 13, 2, 15, 2, 20, 11, 4, 2, 5, 296, 12, 2, 5, 15, 421, 128, 74, 233, 334, 207, 126, 224, 12, 562, 298, 2, 2, 7, 2, 5, 516, 988, 43, 8, 79, 120, 15, 595, 13, 784, 25, 2, 18, 165, 170, 143, 19, 14, 5, 2, 6, 226, 251, 7, 61, 113])]\n",
            "[1 0 0 1 0]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4KQaN7dHLjbE"
      },
      "source": [
        "## 멀티-핫 인코딩\n",
        "\n",
        "[1 3 5] 와 같은 데이터의 경우 1, 3, 5 인덱스만 1이고 나머지는 0으로 인코딩합니다.\n",
        "\n",
        "```\n",
        "[1 3 5] -> [0 1 0 1 0 1 0 0 0 0]   # WORD 길이가 10\n",
        "```\n",
        "\n",
        "본 예의 경우 NUM_WORDS가 1000이다.\n",
        "\n",
        "1개의 데이터는 길이 1000의 벡터이며, 포함된 word 인덱스만 1인 값으로 인코딩 된다."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QpzE4iqZtJly"
      },
      "source": [
        "def multi_hot_sequences(sequences, dimension):\n",
        "    # 0으로 채워진 (25000, 1000) 크기의 행렬. 25000은 데이터 갯수, 1000은 NUM_WORDS\n",
        "    results = np.zeros((len(sequences), dimension))\n",
        "    for i, word_indices in enumerate(sequences):\n",
        "        results[i, word_indices] = 1.0  # results[i]의 특정 인덱스만 1로 설정합니다\n",
        "    return results\n",
        "\n",
        "\n",
        "train_x = multi_hot_sequences(raw_train_x, dimension=NUM_WORDS)\n",
        "test_x = multi_hot_sequences(raw_test_x, dimension=NUM_WORDS)\n",
        "train_y = raw_train_y\n",
        "test_y = raw_test_y"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 146
        },
        "id": "pLsndPYqM8db",
        "outputId": "4797ab79-3d06-428b-be32-8801bd8bf3b4"
      },
      "source": [
        "print(raw_train_x.shape)\n",
        "print(raw_train_x[0])\n",
        "print(train_x.shape)\n",
        "print(train_x[0][:10])\n",
        "print(train_x[0][10:20])\n",
        "print(train_x[0][20:30])\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(25000,)\n",
            "[1, 14, 22, 16, 43, 530, 973, 2, 2, 65, 458, 2, 66, 2, 4, 173, 36, 256, 5, 25, 100, 43, 838, 112, 50, 670, 2, 9, 35, 480, 284, 5, 150, 4, 172, 112, 167, 2, 336, 385, 39, 4, 172, 2, 2, 17, 546, 38, 13, 447, 4, 192, 50, 16, 6, 147, 2, 19, 14, 22, 4, 2, 2, 469, 4, 22, 71, 87, 12, 16, 43, 530, 38, 76, 15, 13, 2, 4, 22, 17, 515, 17, 12, 16, 626, 18, 2, 5, 62, 386, 12, 8, 316, 8, 106, 5, 4, 2, 2, 16, 480, 66, 2, 33, 4, 130, 12, 16, 38, 619, 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, 723, 36, 71, 43, 530, 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",
            "(25000, 1000)\n",
            "[0. 1. 1. 0. 1. 1. 1. 1. 1. 1.]\n",
            "[0. 0. 1. 1. 1. 1. 1. 1. 1. 1.]\n",
            "[0. 1. 1. 0. 0. 1. 1. 0. 1. 0.]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qmppWbxbPj2-"
      },
      "source": [
        "모델이 너무 크면 패턴의 '학습'이 아닌 개별 데이터를 '기억'한다.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3b4k2sqUWvMo"
      },
      "source": [
        "# 모델 크기별 과적합"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L-DGRBbGxI6G"
      },
      "source": [
        "### 작은 모델"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jksi-XtaxDAh",
        "outputId": "7b09d358-f6d6-4efb-b4bd-f536002b5319"
      },
      "source": [
        "small_model = keras.Sequential([\n",
        "    keras.layers.Dense(4, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dense(4, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "small_model.compile(optimizer='adam', loss='binary_crossentropy')\n",
        "\n",
        "small_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_14\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_42 (Dense)             (None, 4)                 4004      \n",
            "_________________________________________________________________\n",
            "dense_43 (Dense)             (None, 4)                 20        \n",
            "_________________________________________________________________\n",
            "dense_44 (Dense)             (None, 1)                 5         \n",
            "=================================================================\n",
            "Total params: 4,029\n",
            "Trainable params: 4,029\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 763
        },
        "id": "Ofn1AwDhx-Fe",
        "outputId": "cea46780-1f40-4f42-bedf-1bf556e6212f"
      },
      "source": [
        "small_history = small_model.fit(train_x,\n",
        "                                    train_y,\n",
        "                                    epochs=20,\n",
        "                                    batch_size=512,\n",
        "                                    validation_split=0.1,\n",
        "                                    verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.6414 - val_loss: 0.5517\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.4776 - val_loss: 0.4257\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.3889 - val_loss: 0.3760\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.3477 - val_loss: 0.3536\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.3270 - val_loss: 0.3441\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.3149 - val_loss: 0.3386\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.3061 - val_loss: 0.3362\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.3002 - val_loss: 0.3372\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.2969 - val_loss: 0.3385\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.2916 - val_loss: 0.3387\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.2884 - val_loss: 0.3409\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.2861 - val_loss: 0.3417\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.2824 - val_loss: 0.3427\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.2814 - val_loss: 0.3465\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.2786 - val_loss: 0.3563\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.2764 - val_loss: 0.3512\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.2739 - val_loss: 0.3522\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.2737 - val_loss: 0.3539\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.2711 - val_loss: 0.3558\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.2699 - val_loss: 0.3587\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_ReKHdC2EgVu"
      },
      "source": [
        "### 기준 모델"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QKgdXPx9usBa",
        "outputId": "32603b74-dcc0-46a1-fca7-eeeb16ef0a02"
      },
      "source": [
        "baseline_model = keras.Sequential([\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dense(16, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "baseline_model.compile(optimizer='adam', loss='binary_crossentropy')\n",
        "\n",
        "baseline_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_15\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_45 (Dense)             (None, 16)                16016     \n",
            "_________________________________________________________________\n",
            "dense_46 (Dense)             (None, 16)                272       \n",
            "_________________________________________________________________\n",
            "dense_47 (Dense)             (None, 1)                 17        \n",
            "=================================================================\n",
            "Total params: 16,305\n",
            "Trainable params: 16,305\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LqG3MXF5xSjR",
        "outputId": "63d25f72-56ae-4f1d-e923-0d48533a59b8"
      },
      "source": [
        "baseline_history = baseline_model.fit(train_x,\n",
        "                                      train_y,\n",
        "                                      epochs=20,\n",
        "                                      batch_size=512,\n",
        "                                      validation_split=0.1,\n",
        "                                      verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.5740 - val_loss: 0.4303\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.3679 - val_loss: 0.3529\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.3205 - val_loss: 0.3392\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.3069 - val_loss: 0.3387\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.2984 - val_loss: 0.3405\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.2939 - val_loss: 0.3436\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.2890 - val_loss: 0.3434\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.2831 - val_loss: 0.3436\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.2776 - val_loss: 0.3463\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.2728 - val_loss: 0.3475\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.2702 - val_loss: 0.3500\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.2615 - val_loss: 0.3528\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.2559 - val_loss: 0.3565\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.2501 - val_loss: 0.3615\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.2437 - val_loss: 0.3615\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.2381 - val_loss: 0.3668\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.2310 - val_loss: 0.3645\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.2258 - val_loss: 0.3721\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.2214 - val_loss: 0.3793\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.2129 - val_loss: 0.3806\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vIPuf23FFaVn"
      },
      "source": [
        "### 큰 모델"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ghQwwqwqvQM9",
        "outputId": "01e74c9b-a8e2-4ba3-e8fc-6a5fe37c24c3"
      },
      "source": [
        "big_model = keras.Sequential([\n",
        "    keras.layers.Dense(512, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dense(512, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "big_model.compile(optimizer='adam', loss='binary_crossentropy')\n",
        "\n",
        "big_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_16\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_48 (Dense)             (None, 512)               512512    \n",
            "_________________________________________________________________\n",
            "dense_49 (Dense)             (None, 512)               262656    \n",
            "_________________________________________________________________\n",
            "dense_50 (Dense)             (None, 1)                 513       \n",
            "=================================================================\n",
            "Total params: 775,681\n",
            "Trainable params: 775,681\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "D-d-i5DaYmr7"
      },
      "source": [
        "역시 같은 데이터를 사용해 모델을 훈련합니다:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "U1A99dhqvepf",
        "outputId": "97c5be63-eb5a-4678-aa25-94ca7eff0cb0"
      },
      "source": [
        "big_history = big_model.fit(train_x,\n",
        "                                  train_y,\n",
        "                                  epochs=20,\n",
        "                                  batch_size=512,\n",
        "                                  validation_split=0.1,\n",
        "                                  verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.4308 - val_loss: 0.3383\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.2933 - val_loss: 0.3558\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.2331 - val_loss: 0.3522\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.1264 - val_loss: 0.4118\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.0329 - val_loss: 0.5255\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.0068 - val_loss: 0.6298\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.0017 - val_loss: 0.6833\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 8.1064e-04 - val_loss: 0.7204\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 5.3614e-04 - val_loss: 0.7491\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 3.8376e-04 - val_loss: 0.7762\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 2.8577e-04 - val_loss: 0.7976\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 2.1922e-04 - val_loss: 0.8176\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 1.7192e-04 - val_loss: 0.8366\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 1.3779e-04 - val_loss: 0.8545\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 1.1212e-04 - val_loss: 0.8713\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 9.2847e-05 - val_loss: 0.8871\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 7.7922e-05 - val_loss: 0.9003\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 6.6015e-05 - val_loss: 0.9154\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 5.6546e-05 - val_loss: 0.9273\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 4.8907e-05 - val_loss: 0.9393\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Fy3CMUZpzH3d"
      },
      "source": [
        "## 로스 그래프"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 825
        },
        "id": "X7uu67SxRp4V",
        "outputId": "4d8a86a7-9d47-46b2-a5f7-57e9fb218dfd"
      },
      "source": [
        "print(\"small\")\n",
        "plt.plot(small_history.history['loss'])\n",
        "plt.plot(small_history.history['val_loss'])\n",
        "plt.ylim((-0.1, 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"base\")\n",
        "plt.plot(baseline_history.history['loss'])\n",
        "plt.plot(baseline_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"big\")\n",
        "plt.plot(big_history.history['loss'])\n",
        "plt.plot(big_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "small\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAduElEQVR4nO3de5hcdZ3n8fe3qrr6lqSbdDohNzQ3gaAI2Abwxs0RgqPB0XFhWHGQHTYMgXH28VH2cddx13lmV2d0BAGRQQQcV7xrhuHiGFRECKQTIRBjoHORNAlJp5N0kk76UlXf/eOc6q6uru6u0NVd3Sef1/MczuX3q6ovh+JzfudU1Wlzd0REZPKLlbsAEREpDQW6iEhEKNBFRCJCgS4iEhEKdBGRiFCgi4hExIiBbmb3mtleM3txiHYzs9vMrMXMNprZOaUvU0RERlLMCP0+4LJh2pcDS8LpeuDroy9LRESO14iB7u5PAPuH6bICeMADa4F6M5tdqgJFRKQ4iRI8x1xgZ856a7htd35HM7ueYBRPbW3t20477bQSvLyIyIlj/fr1+9y9sVBbKQLdCmwreD8Bd78buBugqanJm5ubS/DyIiInDjP741BtpfiWSyswP2d9HrCrBM8rIiLHoRSBvhq4Jvy2y3lAh7sPutwiIiJja8RLLmb2XeBCYIaZtQJ/B1QAuPtdwMPA5UALcBS4dqyKFRGRoY0Y6O5+1QjtDtxYsopEROR10S9FRUQiQoEuIhIRCnQRkYhQoIuIRIQCXUQkIhToIiIRoUAXEYkIBbqISEQo0EVEIkKBLiISEQp0EZGIUKCLiESEAl1EJCIU6CIiEaFAFxGJCAW6iEhEKNBFRCJCgS4iEhEKdBGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiQgFuohIRCjQRUQiQoEuIhIRCnQRkYhQoIuIRIQCXUQkIhToIiIRoUAXEYmIogLdzC4zsy1m1mJmtxRorzOzfzOz581sk5ldW/pSRURkOCMGupnFgTuA5cBS4CozW5rX7Ubg9+7+VuBC4MtmlixxrSIiMoxiRujLgBZ33+buPcCDwIq8Pg5MNTMDpgD7gVRJKxURkWEVE+hzgZ05663htly3A6cDu4AXgL9x90z+E5nZ9WbWbGbNbW1tr7NkEREppJhAtwLbPG/9UuA5YA5wFnC7mU0b9CD3u929yd2bGhsbj7NUEREZTjGB3grMz1mfRzASz3Ut8GMPtADbgdNKU6KIiBSjmEBfBywxswXhB51XAqvz+rwCXAJgZrOAU4FtpSxURESGlxipg7unzGwV8BgQB+51901mtjJsvwv4AnCfmb1AcInmM+6+bwzrFhGRPCMGOoC7Pww8nLftrpzlXcD7SluaiIgcD/1SVEQkIhToIiIRoUAXEYkIBbqISEQo0EVEIkKBLiISEQp0EZGIUKCLiESEAl1EJCIU6CIiEaFAFxGJCAW6iEhEKNBFRCJCgS4iEhEKdBGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiQgFuohIRCjQRUQiQoEuIhIRCnQRkYhQoIuIRMSkC/Tt+zr5/OpN9KYz5S5FRGRCmYSBfoT7ntrBTza8Wu5SREQmlEkX6BedOpMz59Vx+y9bNEoXEckx6QLdzLj54iW8sv8oP3tuV7nLERGZMCZdoANccvpMzpgzjdsff5mURukiIsAkDXQz4+ZLlrCj/Sirn9coXUQEigx0M7vMzLaYWYuZ3TJEnwvN7Dkz22Rmvy5tmYO9b+ksTp89jdsfbyGd8bF+ORGRCW/EQDezOHAHsBxYClxlZkvz+tQDdwIfdPczgD8vfamD6uLmixezbV8nD23UKF1EpJgR+jKgxd23uXsP8CCwIq/PXwA/dvdXANx9b2nLLOzSM07m1FlTuW3Nyxqli8gJr5hAnwvszFlvDbflehNwkpn9yszWm9k1hZ7IzK43s2Yza25ra3t9FeeIxYybLlnM1rZOHn5h96ifT0RkMism0K3AtvzhcAJ4G/B+4FLgf5rZmwY9yP1ud29y96bGxsbjLraQy988myUzp/C1x18mo1G6iJzAign0VmB+zvo8IP+idSvwqLt3uvs+4AngraUpcXjBKH0JL+05wiMvvjYeLykiMiEVE+jrgCVmtsDMksCVwOq8Pj8D3m1mCTOrAc4FNpe21KG9/y2zWdRYq1G6iJzQRgx0d08Bq4DHCEL6++6+ycxWmtnKsM9m4FFgI/AscI+7vzh2ZQ8Ujxk3XbyEP7x2mJ//XqN0ETkxmXt5RrRNTU3e3NxcsudLZ5w/+cqvqayI8+83vYtYrNClfxGRyc3M1rt7U6G2SflL0ULiMePGixazefchfrF5T7nLEREZd5EJdIAVZ83hjQ013LrmZcp15iEiUi6TM9AzhW/IlYjHuPGixWzadYg1m8flt00iIhPG5Av0bb+CO8+DI4V/mHTF2XOZP72a2x7XKF1ETiyTL9CnnAwHtsOjnynYXBGPseqixWxs7eBXW0b/a1QRkcli8gX6zNPg3Z+CF38EWx4t2OVDZ89jbn21rqWLyAll8gU6wLv+FmYuhX//b9B1aFBzMhFcS39u50GeeHlfGQoUERl/kzPQE0n44Nfg0C5Y878KdvnI2+Yxp66KW3/xkkbpInJCmJyBDjCvCc5dCevugVfWDmpOJmLccNFiNrxykN+2tJehQBGR8TV5Ax3g4v8BdafA6pugt2tQ80eb5jG7ropb12iULiLRN7kDvXIKfOCfYd9L8Jt/GtyciHPDhYtYt+MAT2/VKF1Eom1yBzrA4vfCmVfCk/8MezYNav5o03xmTavk1jUvl6E4EZHxM/kDHeDSf4CquuDSSyY9oKmqIs7KCxbxzPb9rN2mUbqIRFc0Ar22AZZ/CV5dD898Y1DzVctOoXFqJbf+QqN0EYmuaAQ6wJs/DEsuhce/AAd2DGiqqojzX9+zkKe3tfPs9v3lqU9EZIxFJ9DN4P1fBovBQ38Led9qufrcNzBjSiW36Vq6iERUdAIdoH4+vPfzsPVxeP7BAU3VyWCU/mTLPtb/UaN0EYmeaAU6QNN1MP9ceOy/D7oj49XnnUJDbZJb17SUqTgRkbETvUCPxeADt0FP56A7MtYkE/zVexbyxEtt/PR3r5apQBGRsRG9QIdh78h47TvfyHkLp/OpHzzPr7boj2CISHREM9AhuCNj4+mD7shYmYhz9zVNvGnWVG741w387pUDZSxSRKR0ohvow9yRcVpVBfd/Yhkzp1Vy7X3raNl7uExFioiUTnQDHWD+24e8I2Pj1Eq+/YlzqYjH+Ng3n2XXwWNlKlJEpDSiHegw7B0ZT2mo4f5rl3GkK8XHvvkMBzp7ylSkiMjoRT/QR7gj49I507jn403sPHCMa+9bR2d3qgxFioiMXvQDHUa8I+O5Cxu4/aqz2dh6kBu+s4GeVKYMRYqIjM6JEegw7B0ZAd53xsn8nz97C0+81ManfvA8mYz+IIaITC4nTqCPcEdGgP/09lP49GWnsvr5Xfzvh36vv3IkIpNKotwFjKs3fxg2fi+4I+OUmcG62YAuN1ywiPYjPXzzye00Tq3kxosWl6lYEZHjc+KM0CEI7w/cCo2nwo+ug29fAe1b87oYn738dD509lz+8bEtfPfZV8pTq4jIcTqxAh1g2hz4L2vg8n+CVzfAnefBL/9hwFcaYzHjSx85kwtPbeSzP3mBR1/cXcaCRUSKU1Sgm9llZrbFzFrM7JZh+r3dzNJm9pHSlTgGYnFY9lewqhmWroBffzEI9pZf9HWpiMe48+pzOGt+PTd/9zme2rqvjAWLiIxsxEA3szhwB7AcWApcZWZLh+j3ReCxUhc5ZqbOgg/fA9f8LAj5f/0wfP/jwe0CCO7OeO9fvp03NNRw/QPrefHVjjIXLCIytGJG6MuAFnff5u49wIPAigL9bgJ+BEy+WxguvBBueAou+ixseQRuXwZrvw7pFPU1SR64bhnTqhL85beeZce+znJXKyJSUDGBPhfYmbPeGm7rY2ZzgQ8Bdw33RGZ2vZk1m1lzW1vbcF3HX6ISLvg03LgWTjkXHr0F/uVCaG1mdl01D1x3LumM87F7n2Hvoa4Rn05EZLwVE+hWYFv+F7S/CnzG3Qf/Yif3Qe53u3uTuzc1NjYWWeI4m74Qrv4h/Pn90LkP7nkv/NsnWTy1l29du4z2Iz1cc++zvLRHd2gUkYmlmEBvBebnrM8DduX1aQIeNLMdwEeAO83silIUWBZmcMYVsGodnPfXsOF++FoTZ7U/wjf+8zm0HjjGpV99gpu/+zu2th0pd7UiIgDYSL+GNLME8BJwCfAqsA74C3cffFOUoP99wEPu/sPhnrepqcmbm5tfT83jb/fG4A9ltK6DN76bjov/L9/4fYL7ntpBV2+aK86ey80XL+GNM2rLXamIRJyZrXf3pkJtI47Q3T0FrCL49spm4PvuvsnMVprZytKWOkHNPhM+8XP406/Cay9Q96138+nWm2h+5zr+7qxOHnuhlUu+8ms+/cPn2bn/aLmrFZET1Igj9LEyqUbouY60BX8wo+UXsGsDeIZMZR1/qD6b/9e+hCfSb+Fdbz+HVRctZk59dbmrFZGIGW6ErkAfjaP7YfuvoWUNbH0cDr0KwFafw5OZM4kvuYQ/ef+HmdXQUOZCRSQqFOjjwT34Ixota+ja8h/E/vhbkt5Njyd4ddpbaTxrOVOWXgqz3gyxE++OCxIRqR44vBviyeB21BXVg25wF1mZDGR6Id0T3CokdWzwPNUNvccg1TX8fPF7gy9evA7DBfqJdbfFsWQW3PSr8VSqzv9r6O1i76Zf8fsnf8qsvb9lym/+Hn7z92RqZhBbeCE0LIa6eeE0H+rmBv9ziJRT1yHoaIWOnXDwlXC+M5h3tMLh1xjwreVYRRDsI071OcvTwGLB3yXIpMDTQVh6OtiWnWdS4JmB23Lb0r1hwPbmLKdytvXk9UuF27LL3cF6tm92OdU9eFu6O3iu0UhUBVNFNUxfMLrnGoJG6ONg+75O7nt0LUc2/5yLEi/wnsqXmdrThuV/nb+2MS/k5w1cr208cUZDUhruQSj1dELPkWDefTi4PNjR2h/WB3dCxyvQlXd7i3gSps2F+vnB3+atnx/c4C6TDvqONKXK9MfXY4mg9lgFxBPhPJxiFUFbPJwnkuF6zrb4UNuyj6mCiipIVA8xD4M7O49XluzMXJdcJoiWvYe5dU0LD23cRcJTLKo8xMWzu1k2/Sin1xxiZqYNO9Ta/z9ab95tBuKVQbhPmwM104NRT/VJUF0fLtcP3lY5TZd4Rss9HOn1hKO3nnC0GI4qPROOMHPXc9t94HrfSDO7nsobfaYLbM/kjGZTwWl7NqS7jwwM7L7pSDANN7JMTg3Den7ePAzv2pmje/+kuoNRf1/IHwwPGg4WD+6hNGAeC+eJAtty+sYSeeGcE9qxRKQHPgr0CWbv4S7WbtvP01vbeXrrPna0B191PKmmgnMXNPCOxQ2cv2A6i6elseypbvY0uKM1uHnYsQPB1HUwCJihWCwI9dyQz177TFQGB4lEzhSvDE8Nw1FIdjTSt14ZLPeFWLrAKXER2zOp/sDqW85f7y3Qng05p+/Uf9Ayhddzl/tOrbMh3R1cHy447+5/vokkOQWStTnTlJz5UG21UDkVpp4chHd1fbn/LeQ4KdAnuF0HjwXhvq2dp7e28+rB4DR1xpRKzl/UwPkLG3jHogbe0FCD5Y883IPRWtfBMOQPhssH+wM/f7mrIwipVFd/YE200LLcUVh2xBZOFu8fgfXtDxu4nN/Wtx4u951qVxaYV4btlf2n2gO2VeTUEQtHjrH+qW89t93y1rMjzbxRZ99INJY3Sk0MHKkmqnXmdYJSoE8i7s7O/cd4ets+nt7azlNb29l7uBuA2XVVnL+wgfMWNvCmk6eyYEYtddUVpXrh/g9/sgGf6gpHsF3BaLVvvTsvmAqcEg+5PZYX0AUCO5YT2CIygAJ9EnN3tu3rDC/PtLN2Wzvtnf2XWGZMSbJgRm04TWHBjFoWNdZySkMNlYl4GSsXkbGgQI+QTMbZ3t7JtrZOtrUdYfu+Trbt62T7vk7awpE8QMxg7knVLJgxhYUzalnYWNsX/HPqqonFNAIWmYz0PfQIicWMRY1TWNQ4BZg1oO1QVy87wnDf1hbO9x1h/Y79dPb039k4mYgxu66KWdOqmF1Xxcl1VZwcLgfbqmmcWklcoS8yqSjQI2RaVQVnzqvnzHn1A7a7O22Hu9kWBv2O9k52d3Sxp6OLDa8cYE9HNz3pzIDHxAxmTu0P+5PD4M+G/vTaJPU1FdRXJ0km9OGcyESgQD8BmBkzp1Uxc1oV5y0cfF8Zd2d/Z08Q8oe6Bsxf6+iipe0IT7bs40h34e8z1ybj1NckOam2gpNqksFyTQX11RV924Nt2e1JairjVMR1IBApJQW6YGY0TKmkYUolb55bN2S/w1297DnUxWsd3Rw42sPBoz0cONobLvfPd+4/yoGjvRzq6mW4j2gSMaO6Ik51Mpwq+uc1yThVFTnLOdurkwmmVSWYVl1BXd6kg4ScyBToUrSpVRVMrapg8cypRfVPZ5yOY719wZ87P9aT5lhvOPUMnncc6x3UpzuVGfE1a5Jx6qormFYVBPzg0E9QV1NBdUWi/yBSEac6GaMy0X9AqaqI6zMEmXQU6DJm4jFjem2S6bXJkjxfOuN09abp7ElxuCtFx7FeOo71ciicdxzt7duWnVoPHGXz7qDvUJeMhpJMxPoCv6oiFpwxhIFfmYiRTAQHgWA+eL1yyO3BAaQmmaA2GRxYaiuD5x30wzGR46BAl0kjHjNqKxPUViYo8iRhgFQ6w6HwQJAd/Xflnhn0punuOyPIFGzPrh/uStGTytCdSofzTP88PfKZRCFmhJeVEn0BX1uZoCYZXGqqTSaoqQzaEzEjHjNiFsz7lyEeixG3YH/FYkbc+ufZvslEcICqDOdVFcGBpqoiRlUiTmW4rrOUyUWBLieMRDxW0jOGoWQyTk86CPbu3uw83bfenQoOFke7UxztSXO0J0VnT/96Z7gt23a4K8XeQ9105mxLpZ20+7CfUZRCRdzCgM+GfxD0lRUxkvHY4LOTcFsyMbAtGR/6bCX7XNmDSGXOYysTcSripjOXIinQRUosFjOqYsF1eKrG9rXcnXQmCPdMBtLupMOwT2ecTLY9b7k7PLvo7s3QlUrT1Rusd/UGB5+uVIau3uBzi67e/vbu3mC9Jx0cmA53pWgPz0pyz1h6wm296dEfcczoOyBUVgQHjeyZg1lwd55s4Fv4j767+JiF7eE6Fjxf7gEl52ylMueAlT1jyfbJ7VcR1pA/JcIzpUQ8ZzkWIx7vP0NKxIzq8EP/UlOgi0xilg2PchcyhNyzldxLU9nw785dDs9eci9jdaeCA0x37noqjXtwMHP6b6IZLHv/LeYcnP6zGM9Z7033nykdONoz8ACX83pjZeUFi7hl+Wklf96J+j4QkQgYcLYyybgHB6Pcs5PsGUs646TCs51UJkMmA6lMpu8MKNuecQ8uj4VnUamMk05nhv168Ggo0EVECjCz8HJLHCjRXU3HmH6FISISEQp0EZGIUKCLiESEAl1EJCIU6CIiEaFAFxGJCAW6iEhEKNBFRCKiqEA3s8vMbIuZtZjZLQXarzazjeH0lJm9tfSliojIcEYMdDOLA3cAy4GlwFVmtjSv23bgAnc/E/gCcHepCxURkeEVM0JfBrS4+zZ37wEeBFbkdnD3p9z9QLi6FphX2jJFRGQkxQT6XGBnznpruG0o1wGPFGows+vNrNnMmtva2oqvUkRERlRMoBe6s3zBmxyb2UUEgf6ZQu3ufre7N7l7U2NjY/FViojIiIq522IrMD9nfR6wK7+TmZ0J3AMsd/f20pQnIiLFKmaEvg5YYmYLzCwJXAmszu1gZqcAPwY+5u4vlb5MEREZyYgjdHdPmdkq4DEgDtzr7pvMbGXYfhfwOaABuDP8U1Apd28au7JFRCSf+Vj/ldkhNDU1eXNzc1leW0RksjKz9UMNmPVLURGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiQgFuohIRCjQRUQiQoEuIhIRCnQRkYhQoIuIRIQCXUQkIhToIiIRoUAXEYkIBbqISEQo0EVEIkKBLiISEQp0EZGIUKCLiESEAl1EJCIU6CIiEaFAFxGJCAW6iEhEKNBFRCJCgS4iEhEKdBGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiQgFuohIRCjQRUQioqhAN7PLzGyLmbWY2S0F2s3MbgvbN5rZOaUvVUREhjNioJtZHLgDWA4sBa4ys6V53ZYDS8LpeuDrJa5TRERGUMwIfRnQ4u7b3L0HeBBYkddnBfCAB9YC9WY2u8S1iojIMIoJ9LnAzpz11nDb8fbBzK43s2Yza25razveWkVEZBjFBLoV2Oavow/ufre7N7l7U2NjYzH1iYhIkYoJ9FZgfs76PGDX6+gjIiJjqJhAXwcsMbMFZpYErgRW5/VZDVwTftvlPKDD3XeXuFYRERlGYqQO7p4ys1XAY0AcuNfdN5nZyrD9LuBh4HKgBTgKXDt2JYuISCEjBjqAuz9MENq52+7KWXbgxtKWJiIix0O/FBURiQgFuohIRCjQRUQiQoEuIhIRCnQRkYhQoIuIRIQCXUQkIhToIiIRoUAXEYkIBbqISEQo0EVEIkKBLiISEQp0EZGIUKCLiESEBXe+LcMLm7UBf3ydD58B7CthOaU20euDiV+j6hsd1Tc6E7m+N7h7wb/hWbZAHw0za3b3pnLXMZSJXh9M/BpV3+iovtGZ6PUNRZdcREQiQoEuIhIRkzXQ7y53ASOY6PXBxK9R9Y2O6hudiV5fQZPyGrqIiAw2WUfoIiKSR4EuIhIREzrQzewyM9tiZi1mdkuBdjOz28L2jWZ2zjjWNt/Mfmlmm81sk5n9TYE+F5pZh5k9F06fG6/6wtffYWYvhK/dXKC9nPvv1Jz98pyZHTKzT+b1Gff9Z2b3mtleM3sxZ9t0M/sPM3s5nJ80xGOHfb+OYX3/aGZ/CP8b/sTM6od47LDvhzGs7/Nm9mrOf8fLh3hsufbf93Jq22Fmzw3x2DHff6Pm7hNyAuLAVmAhkASeB5bm9bkceAQw4DzgmXGsbzZwTrg8FXipQH0XAg+VcR/uAGYM0162/Vfgv/VrBD+YKOv+A94DnAO8mLPtS8At4fItwBeH+HcY9v06hvW9D0iEy18sVF8x74cxrO/zwKeKeA+UZf/ltX8Z+Fy59t9op4k8Ql8GtLj7NnfvAR4EVuT1WQE84IG1QL2ZzR6P4tx9t7tvCJcPA5uBuePx2iVUtv2X5xJgq7u/3l8Ol4y7PwHsz9u8Arg/XL4fuKLAQ4t5v45Jfe7+c3dPhatrgXmlft1iDbH/ilG2/ZdlZgZ8FPhuqV93vEzkQJ8L7MxZb2VwYBbTZ8yZ2RuBs4FnCjSfb2bPm9kjZnbG+FaGAz83s/Vmdn2B9gmx/4ArGfp/onLuv6xZ7r4bggM5MLNAn4myLz9BcNZVyEjvh7G0KrwkdO8Ql6wmwv57N7DH3V8eor2c+68oEznQrcC2/O9YFtNnTJnZFOBHwCfd/VBe8waCywhvBb4G/HQ8awPe6e7nAMuBG83sPXntE2H/JYEPAj8o0Fzu/Xc8JsK+/CyQAr4zRJeR3g9j5evAIuAsYDfBZY18Zd9/wFUMPzov1/4r2kQO9FZgfs76PGDX6+gzZsysgiDMv+PuP85vd/dD7n4kXH4YqDCzGeNVn7vvCud7gZ8QnNbmKuv+Cy0HNrj7nvyGcu+/HHuyl6LC+d4Cfcr9Xvw48KfA1R5e8M1XxPthTLj7HndPu3sG+JchXrfc+y8B/BnwvaH6lGv/HY+JHOjrgCVmtiAcxV0JrM7rsxq4Jvy2xnlAR/bUeKyF19u+CWx2968M0efksB9mtoxgf7ePU321ZjY1u0zwwdmLed3Ktv9yDDkqKuf+y7Ma+Hi4/HHgZwX6FPN+HRNmdhnwGeCD7n50iD7FvB/Gqr7cz2U+NMTrlm3/hd4L/MHdWws1lnP/HZdyfyo73ETwLYyXCD79/my4bSWwMlw24I6w/QWgaRxrexfBKeFG4LlwujyvvlXAJoJP7NcC7xjH+haGr/t8WMOE2n/h69cQBHRdzray7j+Cg8tuoJdg1Hgd0ACsAV4O59PDvnOAh4d7v45TfS0E15+z78O78usb6v0wTvV9O3x/bSQI6dkTaf+F2+/Lvu9y+o77/hvtpJ/+i4hExES+5CIiIsdBgS4iEhEKdBGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiYj/D+qotN8PVYVXAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "base\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "big\n"
          ],
          "name": "stdout"
        },
        {
          "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": "QBnYskh3XZ86"
      },
      "source": [
        "모델이 클수록 오버피팅 발생이 일찍 온다"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ASdv7nsgEFhx"
      },
      "source": [
        "# 과적합 처리"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4rHoVWcswFLa"
      },
      "source": [
        "## weight Regularization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HFGmcwduwVyQ",
        "outputId": "1a22513d-4f3a-41e3-c052-cb14cd964288"
      },
      "source": [
        "ㅣ2_baseline_model = keras.Sequential([\n",
        "    # keras.layers.Dense(16, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(NUM_WORDS,), kernel_regularizer=keras.regularizers.l2(0.001)),\n",
        "    # keras.layers.Dense(16, activation='relu'),\n",
        "    keras.layers.Dense(16, activation='relu', kernel_regularizer=keras.regularizers.l2(0.001)),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "ㅣ2_baseline_model.compile(optimizer='adam', loss='binary_crossentropy')\n",
        "\n",
        "ㅣ2_baseline_model.summary()\n",
        "\n",
        "l2_baseline_history = ㅣ2_baseline_model.fit(train_x,\n",
        "                                    train_y,\n",
        "                                    epochs=20,\n",
        "                                    batch_size=512,\n",
        "                                    validation_split=0.1,\n",
        "                                    verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_17\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_51 (Dense)             (None, 16)                16016     \n",
            "_________________________________________________________________\n",
            "dense_52 (Dense)             (None, 16)                272       \n",
            "_________________________________________________________________\n",
            "dense_53 (Dense)             (None, 1)                 17        \n",
            "=================================================================\n",
            "Total params: 16,305\n",
            "Trainable params: 16,305\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.6060 - val_loss: 0.4783\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.4143 - val_loss: 0.4005\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.3655 - val_loss: 0.3838\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.3495 - val_loss: 0.3757\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.3418 - val_loss: 0.3739\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.3361 - val_loss: 0.3725\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.3327 - val_loss: 0.3718\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.3286 - val_loss: 0.3772\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.3289 - val_loss: 0.3756\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.3256 - val_loss: 0.3682\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.3212 - val_loss: 0.3705\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.3191 - val_loss: 0.3704\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.3173 - val_loss: 0.3692\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.3120 - val_loss: 0.3706\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.3092 - val_loss: 0.3709\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.3084 - val_loss: 0.3730\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.3024 - val_loss: 0.3699\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.2995 - val_loss: 0.3702\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.2957 - val_loss: 0.3735\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.2939 - val_loss: 0.3766\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "utsh4INTDGR4",
        "scrolled": true,
        "outputId": "66d8da66-8f1e-4e11-853b-a9b16afb76b8"
      },
      "source": [
        "ㅣ2_big_model = keras.Sequential([\n",
        "    # keras.layers.Dense(512, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dense(512, activation='relu', input_shape=(NUM_WORDS,), kernel_regularizer=keras.regularizers.l2(0.001)),\n",
        "    # keras.layers.Dense(512, activation='relu'),\n",
        "    keras.layers.Dense(512, activation='relu', kernel_regularizer=keras.regularizers.l2(0.001)),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "ㅣ2_big_model.compile(optimizer='adam', loss='binary_crossentropy')\n",
        "\n",
        "ㅣ2_big_model.summary()\n",
        "\n",
        "l2_big_history = ㅣ2_big_model.fit(train_x,\n",
        "                                    train_y,\n",
        "                                    epochs=20,\n",
        "                                    batch_size=512,\n",
        "                                    validation_split=0.1,\n",
        "                                    verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_18\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_54 (Dense)             (None, 512)               512512    \n",
            "_________________________________________________________________\n",
            "dense_55 (Dense)             (None, 512)               262656    \n",
            "_________________________________________________________________\n",
            "dense_56 (Dense)             (None, 1)                 513       \n",
            "=================================================================\n",
            "Total params: 775,681\n",
            "Trainable params: 775,681\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 1.2307 - val_loss: 0.8746\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.7101 - val_loss: 0.6455\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.5375 - val_loss: 0.5722\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.4531 - val_loss: 0.5397\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.3736 - val_loss: 0.5144\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.3066 - val_loss: 0.5358\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.2398 - val_loss: 0.5473\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.2028 - val_loss: 0.5794\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.1751 - val_loss: 0.5636\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.1567 - val_loss: 0.5532\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.1413 - val_loss: 0.5614\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.1304 - val_loss: 0.5431\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.1210 - val_loss: 0.5526\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.1164 - val_loss: 0.5556\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.1134 - val_loss: 0.5559\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.1100 - val_loss: 0.5351\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.1172 - val_loss: 0.6147\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.1162 - val_loss: 0.5796\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.1119 - val_loss: 0.6027\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.1377 - val_loss: 0.5791\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "cQAECSenUBC4",
        "outputId": "ed07c202-4fb8-4b13-8a76-45012405f717"
      },
      "source": [
        "print(\"baseline\")\n",
        "plt.plot(baseline_history.history['loss'])\n",
        "plt.plot(baseline_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"l2_baseline\")\n",
        "plt.plot(l2_baseline_history.history['loss'])\n",
        "plt.plot(l2_baseline_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"big\")\n",
        "plt.plot(big_history.history['loss'])\n",
        "plt.plot(big_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"l2_big\")\n",
        "plt.plot(l2_big_history.history['loss'])\n",
        "plt.plot(l2_big_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "l2_baseline\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "big\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "l2_big\n"
          ],
          "name": "stdout"
        },
        {
          "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": "BV-hZXvBX-VL"
      },
      "source": [
        "발생하던 오버피팅이 완화되었다."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k3V7W79YV4KH"
      },
      "source": [
        "## Drop Out"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "II8XxgNhV9N8",
        "outputId": "f6ac22d7-0f93-4e80-a938-cccbd2747ddb"
      },
      "source": [
        "dropout_baseline_model = keras.Sequential([\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dropout(0.5),    # ADD\n",
        "    keras.layers.Dense(16, activation='relu'),\n",
        "    keras.layers.Dropout(0.5),    # ADDD\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "\n",
        "dropout_baseline_model.compile(optimizer='adam',\n",
        "                 loss='binary_crossentropy',\n",
        "                 metrics=['accuracy', 'binary_crossentropy'])\n",
        "\n",
        "dropout_baseline_history = dropout_baseline_model.fit(train_x,\n",
        "                                    train_y,\n",
        "                                    epochs=20,\n",
        "                                    batch_size=512,\n",
        "                                    validation_split=0.1,\n",
        "                                    verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.6768 - accuracy: 0.5622 - binary_crossentropy: 0.6768 - val_loss: 0.6291 - val_accuracy: 0.7448 - val_binary_crossentropy: 0.6291\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.6007 - accuracy: 0.6713 - binary_crossentropy: 0.6007 - val_loss: 0.5090 - val_accuracy: 0.8196 - val_binary_crossentropy: 0.5090\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.5209 - accuracy: 0.7427 - binary_crossentropy: 0.5209 - val_loss: 0.4296 - val_accuracy: 0.8408 - val_binary_crossentropy: 0.4296\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.4612 - accuracy: 0.7902 - binary_crossentropy: 0.4612 - val_loss: 0.3776 - val_accuracy: 0.8548 - val_binary_crossentropy: 0.3776\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.4300 - accuracy: 0.8100 - binary_crossentropy: 0.4300 - val_loss: 0.3537 - val_accuracy: 0.8568 - val_binary_crossentropy: 0.3537\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.4053 - accuracy: 0.8298 - binary_crossentropy: 0.4053 - val_loss: 0.3466 - val_accuracy: 0.8604 - val_binary_crossentropy: 0.3466\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.3843 - accuracy: 0.8395 - binary_crossentropy: 0.3843 - val_loss: 0.3362 - val_accuracy: 0.8604 - val_binary_crossentropy: 0.3362\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.3707 - accuracy: 0.8469 - binary_crossentropy: 0.3707 - val_loss: 0.3329 - val_accuracy: 0.8604 - val_binary_crossentropy: 0.3329\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.3581 - accuracy: 0.8528 - binary_crossentropy: 0.3581 - val_loss: 0.3343 - val_accuracy: 0.8612 - val_binary_crossentropy: 0.3343\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.3473 - accuracy: 0.8645 - binary_crossentropy: 0.3473 - val_loss: 0.3330 - val_accuracy: 0.8616 - val_binary_crossentropy: 0.3330\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.3409 - accuracy: 0.8624 - binary_crossentropy: 0.3409 - val_loss: 0.3338 - val_accuracy: 0.8600 - val_binary_crossentropy: 0.3338\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.3278 - accuracy: 0.8699 - binary_crossentropy: 0.3278 - val_loss: 0.3358 - val_accuracy: 0.8600 - val_binary_crossentropy: 0.3358\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.3180 - accuracy: 0.8727 - binary_crossentropy: 0.3180 - val_loss: 0.3437 - val_accuracy: 0.8548 - val_binary_crossentropy: 0.3437\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.3195 - accuracy: 0.8720 - binary_crossentropy: 0.3195 - val_loss: 0.3412 - val_accuracy: 0.8580 - val_binary_crossentropy: 0.3412\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.3149 - accuracy: 0.8761 - binary_crossentropy: 0.3149 - val_loss: 0.3462 - val_accuracy: 0.8612 - val_binary_crossentropy: 0.3462\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.3054 - accuracy: 0.8760 - binary_crossentropy: 0.3054 - val_loss: 0.3492 - val_accuracy: 0.8580 - val_binary_crossentropy: 0.3492\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.3010 - accuracy: 0.8758 - binary_crossentropy: 0.3010 - val_loss: 0.3511 - val_accuracy: 0.8568 - val_binary_crossentropy: 0.3511\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.2961 - accuracy: 0.8815 - binary_crossentropy: 0.2961 - val_loss: 0.3551 - val_accuracy: 0.8556 - val_binary_crossentropy: 0.3551\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.2887 - accuracy: 0.8840 - binary_crossentropy: 0.2887 - val_loss: 0.3555 - val_accuracy: 0.8588 - val_binary_crossentropy: 0.3555\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.2893 - accuracy: 0.8832 - binary_crossentropy: 0.2893 - val_loss: 0.3595 - val_accuracy: 0.8572 - val_binary_crossentropy: 0.3595\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0dYQBs5yDpXF",
        "outputId": "88dd4292-b335-4334-adfb-7f902c6ba2ba"
      },
      "source": [
        "dropout_big_model = keras.Sequential([\n",
        "    keras.layers.Dense(512, activation='relu', input_shape=(NUM_WORDS,)),\n",
        "    keras.layers.Dropout(0.5),    # ADD\n",
        "    keras.layers.Dense(512, activation='relu'),\n",
        "    keras.layers.Dropout(0.5),    # ADDD\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "\n",
        "dropout_big_model.compile(optimizer='adam',\n",
        "                 loss='binary_crossentropy',\n",
        "                 metrics=['accuracy', 'binary_crossentropy'])\n",
        "\n",
        "dropout_big_history = dropout_big_model.fit(train_x,\n",
        "                                    train_y,\n",
        "                                    epochs=20,\n",
        "                                    batch_size=512,\n",
        "                                    validation_split=0.1,\n",
        "                                    verbose=2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "44/44 - 1s - loss: 0.4899 - accuracy: 0.7568 - binary_crossentropy: 0.4899 - val_loss: 0.3564 - val_accuracy: 0.8480 - val_binary_crossentropy: 0.3564\n",
            "Epoch 2/20\n",
            "44/44 - 0s - loss: 0.3349 - accuracy: 0.8552 - binary_crossentropy: 0.3349 - val_loss: 0.3341 - val_accuracy: 0.8564 - val_binary_crossentropy: 0.3341\n",
            "Epoch 3/20\n",
            "44/44 - 0s - loss: 0.2956 - accuracy: 0.8749 - binary_crossentropy: 0.2956 - val_loss: 0.3318 - val_accuracy: 0.8664 - val_binary_crossentropy: 0.3318\n",
            "Epoch 4/20\n",
            "44/44 - 0s - loss: 0.2582 - accuracy: 0.8937 - binary_crossentropy: 0.2582 - val_loss: 0.3442 - val_accuracy: 0.8568 - val_binary_crossentropy: 0.3442\n",
            "Epoch 5/20\n",
            "44/44 - 0s - loss: 0.2121 - accuracy: 0.9177 - binary_crossentropy: 0.2121 - val_loss: 0.3588 - val_accuracy: 0.8576 - val_binary_crossentropy: 0.3588\n",
            "Epoch 6/20\n",
            "44/44 - 0s - loss: 0.1655 - accuracy: 0.9376 - binary_crossentropy: 0.1655 - val_loss: 0.4021 - val_accuracy: 0.8548 - val_binary_crossentropy: 0.4021\n",
            "Epoch 7/20\n",
            "44/44 - 0s - loss: 0.1165 - accuracy: 0.9586 - binary_crossentropy: 0.1165 - val_loss: 0.4638 - val_accuracy: 0.8520 - val_binary_crossentropy: 0.4638\n",
            "Epoch 8/20\n",
            "44/44 - 0s - loss: 0.0874 - accuracy: 0.9683 - binary_crossentropy: 0.0874 - val_loss: 0.4975 - val_accuracy: 0.8532 - val_binary_crossentropy: 0.4975\n",
            "Epoch 9/20\n",
            "44/44 - 0s - loss: 0.0613 - accuracy: 0.9793 - binary_crossentropy: 0.0613 - val_loss: 0.5578 - val_accuracy: 0.8424 - val_binary_crossentropy: 0.5578\n",
            "Epoch 10/20\n",
            "44/44 - 0s - loss: 0.0465 - accuracy: 0.9843 - binary_crossentropy: 0.0465 - val_loss: 0.6283 - val_accuracy: 0.8480 - val_binary_crossentropy: 0.6283\n",
            "Epoch 11/20\n",
            "44/44 - 0s - loss: 0.0370 - accuracy: 0.9869 - binary_crossentropy: 0.0370 - val_loss: 0.6580 - val_accuracy: 0.8512 - val_binary_crossentropy: 0.6580\n",
            "Epoch 12/20\n",
            "44/44 - 0s - loss: 0.0361 - accuracy: 0.9875 - binary_crossentropy: 0.0361 - val_loss: 0.6490 - val_accuracy: 0.8524 - val_binary_crossentropy: 0.6490\n",
            "Epoch 13/20\n",
            "44/44 - 0s - loss: 0.0320 - accuracy: 0.9890 - binary_crossentropy: 0.0320 - val_loss: 0.6925 - val_accuracy: 0.8552 - val_binary_crossentropy: 0.6925\n",
            "Epoch 14/20\n",
            "44/44 - 0s - loss: 0.0275 - accuracy: 0.9908 - binary_crossentropy: 0.0275 - val_loss: 0.7112 - val_accuracy: 0.8520 - val_binary_crossentropy: 0.7112\n",
            "Epoch 15/20\n",
            "44/44 - 0s - loss: 0.0252 - accuracy: 0.9913 - binary_crossentropy: 0.0252 - val_loss: 0.7329 - val_accuracy: 0.8500 - val_binary_crossentropy: 0.7329\n",
            "Epoch 16/20\n",
            "44/44 - 0s - loss: 0.0217 - accuracy: 0.9928 - binary_crossentropy: 0.0217 - val_loss: 0.7677 - val_accuracy: 0.8544 - val_binary_crossentropy: 0.7677\n",
            "Epoch 17/20\n",
            "44/44 - 0s - loss: 0.0247 - accuracy: 0.9916 - binary_crossentropy: 0.0247 - val_loss: 0.7451 - val_accuracy: 0.8500 - val_binary_crossentropy: 0.7451\n",
            "Epoch 18/20\n",
            "44/44 - 0s - loss: 0.0236 - accuracy: 0.9917 - binary_crossentropy: 0.0236 - val_loss: 0.7902 - val_accuracy: 0.8496 - val_binary_crossentropy: 0.7902\n",
            "Epoch 19/20\n",
            "44/44 - 0s - loss: 0.0202 - accuracy: 0.9933 - binary_crossentropy: 0.0202 - val_loss: 0.7943 - val_accuracy: 0.8516 - val_binary_crossentropy: 0.7943\n",
            "Epoch 20/20\n",
            "44/44 - 0s - loss: 0.0193 - accuracy: 0.9930 - binary_crossentropy: 0.0193 - val_loss: 0.7972 - val_accuracy: 0.8528 - val_binary_crossentropy: 0.7972\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "9aJPcL5bWOoM",
        "outputId": "0df36e07-8347-4904-91eb-3be909cdd926"
      },
      "source": [
        "print(\"baseline\")\n",
        "plt.plot(baseline_history.history['loss'])\n",
        "plt.plot(baseline_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"dropout_baseline\")\n",
        "plt.plot(dropout_baseline_history.history['loss'])\n",
        "plt.plot(dropout_baseline_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"big\")\n",
        "plt.plot(big_history.history['loss'])\n",
        "plt.plot(big_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()\n",
        "\n",
        "print(\"dropout_big\")\n",
        "plt.plot(dropout_big_history.history['loss'])\n",
        "plt.plot(dropout_big_history.history['val_loss'])\n",
        "plt.ylim((0., 1.))\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "baseline\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAdZUlEQVR4nO3de3Sc9X3n8fd3bhrdZdmyLdlOwIkBGwiYKIa2C2VLLrbTxabN2QPJgSybsy7ZsCc5e3IOtN1Nes62u01z0ks2JJSkhKRNSkLDxSEmsE2T0mzXBJnY2MYxNuYmS7Zl634ZzUj67R/PM9JoPCON7JFm9OjzOuc5z+03M18/Gn/mmefyG3POISIii1+o1AWIiEhxKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgZg10M3vYzM6Y2aE8683Mvmxmx83sZTO7rvhliojIbArZQ38E2DrD+m3ABn/YBXzt4ssSEZG5mjXQnXPPA90zNNkBfNt59gINZtZcrAJFRKQwkSI8xxrg7Yz5dn9ZZ3ZDM9uFtxdPdXX1e6+44ooivLyIyNKxb9++s865plzrihHolmNZzv4EnHMPAQ8BtLa2ura2tiK8vIjI0mFmb+ZbV4yrXNqBdRnza4GOIjyviIjMQTECfTdwl3+1yw1An3PuvMMtIiIyv2Y95GJmfw/cDKwws3bg80AUwDn3ILAH2A4cB4aBu+erWBERyW/WQHfO3THLegd8qmgViYjIBdGdoiIiAaFAFxEJCAW6iEhAKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCQoEuIhIQCnQRkYBQoIuIBIQCXUQkIBToIiIBoUAXEQkIBbqISEAo0EVEAkKBLiISEAp0EZGAUKCLiASEAl1EJCAU6CIiAaFAFxEJCAW6iEhAKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQBQU6Ga21cyOmtlxM7s/x/p6M/uhmR0ws8NmdnfxSxURkZnMGuhmFgYeALYBm4A7zGxTVrNPAa84564Bbga+ZGaxItcqIiIzKGQPfQtw3Dl3wjmXBB4FdmS1cUCtmRlQA3QDY0WtVEREZlRIoK8B3s6Yb/eXZfoKsBHoAA4Cn3bOTWQ/kZntMrM2M2vr6uq6wJJFRCSXQgLdcixzWfMfAvYDLcC1wFfMrO68Bzn3kHOu1TnX2tTUNMdSRURkJoUEejuwLmN+Ld6eeKa7gced5zjwOnBFcUoUEZFCFBLoLwIbzOxS/0Tn7cDurDZvAbcAmNkq4HLgRDELFRGRmUVma+CcGzOze4FngTDwsHPusJnd469/EPgfwCNmdhDvEM19zrmz81i3iIhkmTXQAZxze4A9WcsezJjuAD5Y3NJERGQudKeoiEhAKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCQoEuIhIQCnQRkYBQoIuIBIQCXUQkIBToIiIBoUAXEQkIBbqISEAo0EVEAkKBLiISEAp0EZGAUKCLiASEAl1EJCAU6CIiAaFAFxEJCAW6iEhAKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCoqBAN7OtZnbUzI6b2f152txsZvvN7LCZ/XNxyxQRkdlEZmtgZmHgAeADQDvwopntds69ktGmAfgqsNU595aZrZynekVEJI9C9tC3AMedcyecc0ngUWBHVpuPAo87594CcM6dKW6ZIiIym0ICfQ3wdsZ8u78s02XAMjP7mZntM7O7cj2Rme0yszYza+vq6rqwikVEJKdCAt1yLHNZ8xHgvcCHgQ8B/93MLjvvQc495Jxrdc61NjU1zblYERHJb9Zj6Hh75Osy5tcCHTnanHXODQFDZvY8cA3walGqFBGRWRWyh/4isMHMLjWzGHA7sDurzVPAjWYWMbMq4HrgSHFLFRGRmcy6h+6cGzOze4FngTDwsHPusJnd469/0Dl3xMx+DLwMTADfcM4dms/CRURkOnMu+3D4wmhtbXVtbW1zftzPj53lT/Yc4dH/dAP1VdF5qExEpHyZ2T7nXGuudYvuTtGGqihHOvvZc6iz1KWIiJSVRRfoV7bU8a6map745clSlyIiUlYWXaCbGTuvXcMvXu/mZO9IqcsRESkbiy7QAXZc693X9NR+7aWLiKQtykB/x/Iq3vvOZTz5y5OU6qSuiEi5WZSBDrBz8xpePT3Ikc6BUpciIlIWFm2gf/jqZiIh02EXERHfog30xuoYN1/exFP7Oxif0GEXEZFFG+jgnRw91Z/ghdfPlboUEZGSW9SB/v6Nq6ipiPCkrkkXEVncgV4ZC/OhK1fzzMFTJFLjpS5HRKSkFnWgA9y2eQ0Do2P806/0I0kisrQt+kD/tXctZ2VthQ67iMiSt+gDPRwybr2mhZ8ePUPvcLLU5YiIlMyiD3TwbjJKjTv2HDxV6lJEREomEIF+ZUsd715Zo8MuIrKkBSLQvR4YW/jFG9209wyXuhwRkZIIRKBDZg+M2b9fLSKyNAQm0Nc1VtGqHhhFZAkLTKCDd3L02JlBXunsL3UpIiILLlCBPtUDow67iMjSE6hAX1Yd4+bLV/LU/pPqgVFElpxABTrAzs0tnO4f5YUT6oFRRJaWwAV6ugfGJ3RNuogsMYEL9Hg0zNarVvPjQ+qBUUSWlsAFOkz1wPiTI+qBUUSWjkAG+g3r/R4Y9XujIrKEBDLQwyFjx7Ut/Ew9MIrIEhLIQAevK4DUuONHBztLXYqIyIIIbKBf2VLHBvXAKCJLyOIL9MEueOY+GBudsZmZsXPzGl58o4e3u9UDo4gE3+IL9Dd/Di88CI/dDeOpGZveek0LALsPqCsAEQm+xRfoV94G274IR38ET34SJvJfa76usYr3XbKMJ9QDo4gsAYsv0AGu3wW3fB4OPgZPfwZmCOsd167h+JlBDneoB0YRCbbFGegAN/5XuPGz8NK34dk/yBvqH766mWjYeErXpItIwBUU6Ga21cyOmtlxM7t/hnbvM7NxM/tI8UqcwW/9N7j+k7D3q/DT/5mzybLqGL952Uqe2t+hHhhFJNBmDXQzCwMPANuATcAdZrYpT7svAM8Wu8gZioOt/ws23wnP/xn8/C9zNrtt8xrODIyyVz0wikiAFbKHvgU47pw74ZxLAo8CO3K0+y/AD4CF7UDFDP7dX8FVvwv/+Hn4xdfPa3LLxpXqgVFEAq+QQF8DvJ0x3+4vm2Rma4DbgAdneiIz22VmbWbW1tXVNdda8wuF4ba/hsu3w57Pwv7vTlsdj4bZph4YRSTgCgl0y7Es+2D0XwL3OedmTEvn3EPOuVbnXGtTU1OBJRYoHIWPfBPW3wxPfQoOPzlt9c7NaxgcHeMfj5wu7uuKiJSJQgK9HViXMb8WyL5TpxV41MzeAD4CfNXMdhajwDmJxuH278LaLfCDT8Crz02uumH9clbVVagrABEJrEgBbV4ENpjZpcBJ4Hbgo5kNnHOXpqfN7BHgaefck8Urcw5i1fCx78O3boXv3wkfewwuvYlwyLj1mha++X/foHsoSWN1rCTlicgiM56CwdMwcBoGT3nTqRFv+UTKu7lxPAUTY1NDznUpGPfXb9oB191Z9FJnDXTn3JiZ3Yt39UoYeNg5d9jM7vHXz3jcvCTi9XDnE/DN7fDd2+GuJ2HdFnZuXsPX/+V1fnSwkztveGepqxSRUkolvIAeOA0DnX5on8oad8JwAVfHWRhCEe/QbygMoWie+Yg3To3Myz/JSnVLfGtrq2tra5vfFxk4Bd/cBkPn4D88jVt9NR/8i+epr4zyD5/89fl9bREpjbHkVBgPdHo5kGuc6Dv/saEI1KzyhtrV/rgZaldBzeqpcaxqKqRDEQgt3D2aZrbPOdeaa10hh1wWr9rVcNduL9T/9jbs7j3s3LyGLz57lL/b+yYf3fIOQqFc53xFZM6cg9EBGOmGkR4Y9seZw3A3JHq9PdpIDCJxiFTkH4dzLYt5z9HfkRXS/vTw2fNrC0W9PKhdDcvfDZfcODWfGdRVyxc0nIst2Hvoaedeg4e3QihM/x0/5PeePsf/O3GOLZc28qe/czXrm2oWpg6RhTAx7u19JnphpNcL0oQ/Hun11p3XqV2eHMiVD25ienBnhvdMF7rFaqFyGVTWey83loDxUa8r7LHE1HguLATVK/1wbp4a1zVPn69sXNRBnWmmPfSlEegAp1+BR7ZDRS3u7mf4/qsT/PGPjjA6NsGnb9nArpvWEw0H4w8ui9j4GCQHIDkEo4PeODlw/nSiL0dQ+9OJfvIGNHh7uqFo7nWW6xtrjmXxOqhs8AO60RtX+ePMZenl8QZvj3w2zsF4cnrAj2XPJ7znrW2G6ibvuPQSokBP6/ild/VL1XK45nZ66y7niwdifOeoY2NzPV/43at5z9qGha1JFpeJce+EVmoEUsNZ41zL8qxLDkHSD+nRganpQvdQLewHZoM3jjdMzWdO51oXjc/PtpEFoUDP9NZe+OGnoeso6b2YVLSWg2PrOJBax4p3t/L+37qFyparvON1UhrOeeGWGpkaZ05nXiI2OYznuHRsPEeb1NQeX/o5xxLeVQ85p0f8vUP/defKQhCtgmilN0QqoaLGu8Q2VpsxXeMNFf44Vg0VtVPrMqdj1Xn2piXoFOi5JIe8wzCnD8Kpg4x1HGSi8yAx5+0hTViEUNNlsPpqb1h1Fax+D1QvL13Nac5l7PUNeePkUMZe4DAkhzPWDU8tTw17YTX5ldwPhclwKHAel3F8NT1d4DI3MfX1eTKk/brGRqbG88amTrBFK/3puLfnGqnMWF7hzUfjU23S7TIDenI6e+xPh2MKXymapXuVy0xi1bDufd6AvyEmxtn/8ks8+ePnWDF4lFuGT3P5iecJvfy9qcfVNkPTFd5/bAt5/1HNvGn88eSQMU9mO7w9vbFR73jheMo7OTSeyrEsmbE8Y9lchWN+0FR5gWSh6UELBcyTMW8Zh1b9f1vmGPIvs9BUWMbrILLKD7/4VIBGq/zw9AM3XXc6cMMx/3KxsH+tb2RqfnI636BzJRJMSzfQcwmFufba93HFldfx5Z8c47efP8Gyqih/8uEWPtjYhfl785w9BhPnpvY28ceTQ8Z85l5pZvtwzAuicIU/jnlBFauCcIO/3h8isenz4ZjXbjKgK70PqGglRP1xrGpqOlq15E4ciSxFS/eQSwEOd/Rx3w9e5tDJft6/cRV/vPMqVtfrhJKIlM5Mh1z03XMGV7bU8+R//g3+YPsV/Px4Fx/483/m7/a+yYR++UhEypD20Av05rkhfv/xg/zra+e47h0N3LJxFZta6tjUXMfK2gpMJ71EZAHoKpcicc7x2L52vvJPx3mre3hyeWN1jE3NdWxsrmVjcx2bWup4V1ONblQSkaJToM+DvpEUv+rs50hnP6909nOkc4CjpwdIjk0AEAuHePfKGja11LHRD/tNzXU0VKnbXhG5cLpscR7UV0a5fv1yrl8/dV362PgEJ84OTQv5nx3t4h/2tU+2aamPc0VzHWuXVdLS4A/1cVoaKllZW0FEe/UicoEU6EUUCYe4bFUtl62qZce1Uz+72jUwmhHy/Rw9NUDbG930J6bfdRgOGavr4jT7Ae8NcVrqp6brK6M6Xi8iOSnQF0BTbQVNtU3cdNn031EdSKTo7EvQ0TtCR68/7huho3eEA+29/PjQKZLjE9MeUxULs7o+TkNllNp4lLrKKLXxCHVxf1wZpS5rPr2+KhbWh4FIgCnQS6g27oXyZatqc66fmHCcHRql0w/7k70jdPYlONWXoD+Ronc4ydvdw/QnUvSPjJ0X/tnCIZsM97rKCPWV0cmhrjJKQ2Vs2rLMoTYeUd/xImVOgV7GQiFjZW2clbVxrlnXMGv7RGqc/kSKgcQY/SP+OMd8/0iKPn841Zegb8RbP9MHghnUVkSor/KCf1l1jMaqKI3VFTRW5x7XV0YJ60NAZMEo0AMkHg0Tj4ZZmXuHf0bOORKpicmg7x1OTk73jUz/EOgdSdEzlOT1s4N0DyYZSub+UYOQQUNVjGVVUZZXV7AsI/CXVcW8oTpKQ1WMRn9e3wRELpwCXQAwMypjYSr9Y/RzkUiN0zOcpHto+tAzlOTcUHJy3etnh9j3Zi89w0nG89xtm/khsKwq5oV99dT0sqooy6q9Q0N18Sj1Vd45g5qKiM4PyJKnQJeLFo+Gaa6vpLm+sqD2ExOOgdExevyw7x1OTYZ+erpnOEnPUIr2nmEOnUzRPZycvMY/l3DIvJPB6XMC8fS5gfSJ4unnCyqjYeLRkPetJuJPx7zpaNj04SCLkgJdFlwoZJPhegnVBT3GOcdIapyeYe9wT/oQUH8ifUhobOrwkL+ss2+koPMD59VnU4ev4hEv9CvSHwCRMFWxMCvr4qxpmLq8dE1DJavr47o7WEpKgS6LgplRFYtQFYuwpqGwbwJpzjlGxyamnQsYSY2TSE2QSI17w9gEo+np9PKxzDYTjI556zv7Uhxo7+Xs4PR+6c1gVW3cu3fAD/nscV2lDg3J/FGgS+CZ2eQe96q64nV/nEiNT7uH4GTvyOT40Mk+njt8+rxvBtWxMC0NlSyvSZ8U9k4IN1RFaayeWpY+V1CrcwMyBwp0kQsUj4ZZ31TD+qaanOsnJhznhpJ+6HtBnw79nqEUx84M0jucpGc4lfckcSRkkyeGJ68G8k8SN1Z7Q/pDIT1drRvIliwFusg8CYXMv0u4Ysb7CCYmHAOJMe/E8HCS3uEk3UMpfzx1grh7OMmJs4N0vznzh0AsEvKDf+oKofTef/qbwfLqGMtrKibndb9AMCjQRUosFDLqq7xLMOdykrg/4V0p1D2cpHvQG6fnezIuH+3o7ad7yLuvIBczaPTDfnm1F/Iraiqmhf6KjHW6RLR8KdBFFiGzuV8pNDY+Qe9Iiu6hJOcGk5wbGvXGg6OcHfLG5waTHO7o5+zgKANZncelpb8BTD/uP/1msfR0Y7XXTh8CC0OBLrJERMIhVtRUsKKmAlbN3n50bHwy/M/6YZ/+EPAOBXn3DBw51U/PUJLekRT5fl4hGrbJcwANVVFW1FawsraClbVxmtLTdd58Q2VUdwtfIAW6iORUEZnbDWPjE47+kek3huU7J/BKRz8/60/k7DYiGjZW1Hgh35QV+E01Faysi3sfDNVRXQWURYEuIkURDpl3yWV14b/KNTQ6RtfAKGcGRjkzkJia7vfm23uG+eVbPZwbSuZ8fNi/Sa2hMkpDlXclUEOldz5imf9toL4yOtltRENljPoq74MgiN8CFOgiUjLVFRGqKyJcsmLm8wCp8QnODSY5M5DgTP8oPX7ncemuInpHUvQNpzgzkODV0wP0DqcYHM19DgC8E8E1sanfC8j8TYFav3vp2nj0vOX1ld58Y3WsLH9dTIEuImUvGg6xuj4+p47jUuMTk6HfN+IdAur1exLtz+hSeiDhdRdxqj/BsTNTXU7nuywUvO4hVmX/ulh9nGb/juDm+jiN1bEFPxykQBeRQIpmngSeI+ccw8nxaYHfnxhjIOH1GdTVn+Bkb4LOPv+u4FdOn9d5XEUkNPnTkc31U78d3NxQyYaVNbTMsQuLQhQU6Ga2FfgrIAx8wzn3p1nrPwbc588OAp90zh0oZqEiIgvFzCYPBxXyrcA5767gzt7E5N3AnX1+txB9I/zLsS7ODIxOXgX0ezet5/e3byx63bMGupmFgQeADwDtwItmtts590pGs9eB33TO9ZjZNuAh4PqiVysiUobMbPLbwNVr63O2SY1PcKovQWdfghU1hZ84notC9tC3AMedcycAzOxRYAcwGejOuX/NaL8XWFvMIkVEFrtoOMS6xirWNVbN22sUcpp2DfB2xny7vyyfTwDP5FphZrvMrM3M2rq6ugqvUkREZlVIoOc6TZvz9K+Z/Vu8QL8v13rn3EPOuVbnXGtTU1PhVYqIyKwKOeTSDqzLmF8LdGQ3MrP3AN8AtjnnzhWnPBERKVQhe+gvAhvM7FIziwG3A7szG5jZO4DHgTudc68Wv0wREZnNrHvozrkxM7sXeBbvssWHnXOHzewef/2DwOeA5cBX/Qvpx5xzrfNXtoiIZDOXr3u0edba2ura2tpK8toiIouVme3Lt8Ncfp0RiIjIBVGgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCQoEuIhIQCnQRkYBQoIuIBIQCXUQkIBToIiIBoUAXEQkIBbqISEAo0EVEAkKBLiISEAp0EZGAUKCLiASEAl1EJCAU6CIiAaFAFxEJCAW6iEhAKNBFRAJCgS4iEhAKdBGRgFCgi4gEhAJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCQoEuIhIQCnQRkYBQoIuIBIQCXUQkIAoKdDPbamZHzey4md2fY72Z2Zf99S+b2XXFL1VERGYya6CbWRh4ANgGbALuMLNNWc22ARv8YRfwtSLXKSIisyhkD30LcNw5d8I5lwQeBXZktdkBfNt59gINZtZc5FpFRGQGkQLarAHezphvB64voM0aoDOzkZntwtuDBxg0s6NzqnbKCuDsBT52IZR7fVD+Naq+i6P6Lk451/fOfCsKCXTLscxdQBuccw8BDxXwmjMXZNbmnGu92OeZL+VeH5R/jarv4qi+i1Pu9eVTyCGXdmBdxvxaoOMC2oiIyDwqJNBfBDaY2aVmFgNuB3ZntdkN3OVf7XID0Oec68x+IhERmT+zHnJxzo2Z2b3As0AYeNg5d9jM7vHXPwjsAbYDx4Fh4O75KxkowmGbeVbu9UH516j6Lo7quzjlXl9O5tx5h7pFRGQR0p2iIiIBoUAXEQmIsg70cu5ywMzWmdlPzeyImR02s0/naHOzmfWZ2X5/+NxC1ee//htmdtB/7bYc60u5/S7P2C77zazfzD6T1WbBt5+ZPWxmZ8zsUMayRjP7P2Z2zB8vy/PYGd+v81jfF83sV/7f8Akza8jz2BnfD/NY3x+Z2cmMv+P2PI8t1fb7XkZtb5jZ/jyPnfftd9Gcc2U54J2AfQ1YD8SAA8CmrDbbgWfwroO/AXhhAetrBq7zp2uBV3PUdzPwdAm34RvAihnWl2z75fhbnwLeWertB9wEXAccylj2Z8D9/vT9wBfy/BtmfL/OY30fBCL+9Bdy1VfI+2Ee6/sj4LMFvAdKsv2y1n8J+Fyptt/FDuW8h17WXQ445zqdcy/50wPAEby7YxeTcumy4RbgNefcmyV47Wmcc88D3VmLdwDf8qe/BezM8dBC3q/zUp9z7jnn3Jg/uxfvPpCSyLP9ClGy7ZdmZgb8e+Dvi/26C6WcAz1fdwJzbTPvzOwSYDPwQo7Vv2ZmB8zsGTO7cmErwwHPmdk+v9uFbGWx/fDubcj3n6iU2y9tlfPvq/DHK3O0KZdt+R/xvnXlMtv7YT7d6x8SejjPIaty2H43Aqedc8fyrC/l9itIOQd60bocmE9mVgP8APiMc64/a/VLeIcRrgH+N/DkQtYG/IZz7jq83jA/ZWY3Za0vh+0XA24FHsuxutTbby7KYVv+ITAGfCdPk9neD/Pla8C7gGvx+nf6Uo42Jd9+wB3MvHdequ1XsHIO9LLvcsDMonhh/h3n3OPZ651z/c65QX96DxA1sxULVZ9zrsMfnwGewPtam6kcumzYBrzknDudvaLU2y/D6fShKH98JkebUr8XPw78NvAx5x/wzVbA+2FeOOdOO+fGnXMTwNfzvG6pt18E+B3ge/nalGr7zUU5B3pZdzngH2/7G+CIc+7P87RZ7bfDzLbgbe9zC1RftZnVpqfxTpwdympWDl025N0rKuX2y7Ib+Lg//XHgqRxtCnm/zgsz2wrcB9zqnBvO06aQ98N81Zd5Xua2PK9bsu3nez/wK+dce66Vpdx+c1Lqs7IzDXhXYbyKd/b7D/1l9wD3+NOG9+MbrwEHgdYFrO3f4H0lfBnY7w/bs+q7FziMd8Z+L/DrC1jfev91D/g1lNX281+/Ci+g6zOWlXT74X24dAIpvL3GTwDLgZ8Ax/xxo9+2Bdgz0/t1geo7jnf8Of0+fDC7vnzvhwWq72/999fLeCHdXE7bz1/+SPp9l9F2wbffxQ669V9EJCDK+ZCLiIjMgQJdRCQgFOgiIgGhQBcRCQgFuohIQCjQRUQCQoEuIhIQ/x8W82II1bkH2QAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dropout_baseline\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "big\n"
          ],
          "name": "stdout"
        },
        {
          "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"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "dropout_big\n"
          ],
          "name": "stdout"
        },
        {
          "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": {
        "id": "qhsGpCKMp4Y-"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}