10sequence_modeling.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "colab": {
      "name": "sequence-modeling.ipynb",
      "provenance": [],
      "include_colab_link": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/hikmatfarhat-ndu/CSC645/blob/master/10sequence_modeling.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "13NLGKdqH92v",
        "outputId": "1b98f85f-fdc9-4ea0-ae5c-a361d575da99",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras.datasets import imdb\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "from tensorflow.keras.layers import Dropout,Embedding,LSTM,Dense\n",
        "print(tf.__version__)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.3.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I_xBeZe-CHhf"
      },
      "source": [
        "(x_train,y_train),(x_test,y_test)=imdb.load_data()"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H6Binkr-ETFn",
        "outputId": "c7f787f3-7809-4f23-f152-a69c77d9776f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "for i in range(5):\n",
        " print(len(x_train[i]),x_train[i][0:10])\n"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "218 [1, 14, 22, 16, 43, 530, 973, 1622, 1385, 65]\n",
            "189 [1, 194, 1153, 194, 8255, 78, 228, 5, 6, 1463]\n",
            "141 [1, 14, 47, 8, 30, 31, 7, 4, 249, 108]\n",
            "550 [1, 4, 18609, 16085, 33, 2804, 4, 2040, 432, 111]\n",
            "147 [1, 249, 1323, 7, 61, 113, 10, 10, 13, 1637]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0nH0tWhEH95A"
      },
      "source": [
        "#### START HERE Load and transform the IMDB review sentiment dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eR-uHxhiH95B"
      },
      "source": [
        "# load and preprocess the IMDB dataset\n",
        "num_words=5000\n",
        "maxlen=300\n",
        "index_from=3\n",
        "(x_train, y_train), (x_test, y_test) = imdb.load_data(path='imdb.npz',num_words=num_words,maxlen=maxlen)\n",
        "\n"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nH8-qLFYBqhd",
        "outputId": "865d028d-4aad-4129-df3f-0ff47d3691c8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "for i in range(4):\n",
        " print(len(x_train[i]),x_train[i][0:10])\n"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "218 [1, 14, 22, 16, 43, 530, 973, 1622, 1385, 65]\n",
            "189 [1, 194, 1153, 194, 2, 78, 228, 5, 6, 1463]\n",
            "141 [1, 14, 47, 8, 30, 31, 7, 4, 249, 108]\n",
            "147 [1, 249, 1323, 7, 61, 113, 10, 10, 13, 1637]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4tB9kOCvBnlk"
      },
      "source": [
        "x_train = pad_sequences(x_train,maxlen=None,padding='pre',truncating='pre',value=0)\n",
        "    \n",
        "x_test =pad_sequences(x_test,maxlen=None,padding='pre',truncating='pre',value=0)"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aDqR0iivCjgU",
        "outputId": "eacc1200-9a0c-4d6d-e63b-9952debcac43",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "print(x_train.shape)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(19051, 299)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rwd0mwJkH95D"
      },
      "source": [
        "# Load the dataset\n",
        "\n",
        "#(x_train,y_train),(x_test,y_test)=get_and_pad_imdb_dataset(num_words=num_words,maxlen=300)"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gPANcwE1H95H"
      },
      "source": [
        "# A function to get the dataset word index\n",
        "\n",
        "def get_imdb_word_index(num_words=5000, index_from=2):\n",
        "    imdb_word_index = imdb.get_word_index(\n",
        "                                        path='imdb_word_index.json')\n",
        "    imdb_word_index = {key: value + index_from for\n",
        "                       key, value in imdb_word_index.items() if value <= num_words-index_from}\n",
        "    return imdb_word_index"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qvuhek9BH95J"
      },
      "source": [
        "# Get the word index using get_imdb_word_index()\n",
        "imdb_word_index=get_imdb_word_index(num_words=num_words)\n"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aav965QDH95L"
      },
      "source": [
        "#### Create a recurrent neural network model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t9ltGa8pH95L"
      },
      "source": [
        "# Get the maximum index value\n",
        "\n",
        "max_index_value=max(imdb_word_index.values())\n",
        "embedding_dim=256"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2FHM89-4H95O"
      },
      "source": [
        "model=tf.keras.Sequential()\n",
        "model.add(Embedding(input_dim=max_index_value+1,output_dim=embedding_dim,mask_zero=True))\n",
        "model.add(LSTM(units=60))\n",
        "model.add(Dense(units=1,activation='sigmoid'))    \n",
        "\n"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KWPRGm6sH95R"
      },
      "source": [
        "#### Compile and fit the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zotvkXSfH95R"
      },
      "source": [
        "# Compile the model with binary cross-entropy loss\n",
        "\n",
        "model.compile(loss='binary_crossentropy',metrics=['accuracy'],optimizer='adam')"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JgUS_2WIH95T",
        "outputId": "da5427c1-3302-4bee-81db-987123aaed1b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# Fit the model and save its training history\n",
        "history=model.fit(x_train,y_train,epochs=3,batch_size=500,validation_data=(x_test,y_test))\n"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/3\n",
            "39/39 [==============================] - 32s 831ms/step - loss: 0.5515 - accuracy: 0.7220 - val_loss: 0.3585 - val_accuracy: 0.8540\n",
            "Epoch 2/3\n",
            "39/39 [==============================] - 31s 801ms/step - loss: 0.2864 - accuracy: 0.8877 - val_loss: 0.3511 - val_accuracy: 0.8609\n",
            "Epoch 3/3\n",
            "39/39 [==============================] - 32s 812ms/step - loss: 0.2274 - accuracy: 0.9135 - val_loss: 0.2993 - val_accuracy: 0.8790\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jpiwIFlfOo0b",
        "tags": [
          "outputPrepend"
        ],
        "outputId": "d5d7d65b-e7eb-4b86-85d1-111f12a67474",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "model.evaluate(x_test,y_test,verbose=1)\n",
        "print(\"done\")"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "608/608 [==============================] - 34s 55ms/step - loss: 0.2993 - accuracy: 0.8790\n",
            "done\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NoJGrfl3P4py",
        "outputId": "47d60849-4687-4f42-e8ab-175123adee45",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "emb=model.layers[0]\n",
        "lstm=model.layers[1]\n",
        "dense=model.layers[2]\n",
        "y=emb(x_train[0:1])\n",
        "print(x_train[0:1])\n",
        "print(y)\n",
        "z=lstm(y)\n",
        "print(z)\n",
        "logit=dense(z)\n",
        "print(logit)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[   0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
            "     0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
            "     0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
            "     0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
            "     0    0    0    0    0    0    0    0    0    0    0    0    0    0\n",
            "     0    0    0    0    0    0    0    0    0    0    0    1   14   22\n",
            "    16   43  530  973 1622 1385   65  458 4468   66 3941    4  173   36\n",
            "   256    5   25  100   43  838  112   50  670    2    9   35  480  284\n",
            "     5  150    4  172  112  167    2  336  385   39    4  172 4536 1111\n",
            "    17  546   38   13  447    4  192   50   16    6  147 2025   19   14\n",
            "    22    4 1920 4613  469    4   22   71   87   12   16   43  530   38\n",
            "    76   15   13 1247    4   22   17  515   17   12   16  626   18    2\n",
            "     5   62  386   12    8  316    8  106    5    4 2223    2   16  480\n",
            "    66 3785   33    4  130   12   16   38  619    5   25  124   51   36\n",
            "   135   48   25 1415   33    6   22   12  215   28   77   52    5   14\n",
            "   407   16   82    2    8    4  107  117    2   15  256    4    2    7\n",
            "  3766    5  723   36   71   43  530  476   26  400  317   46    7    4\n",
            "     2 1029   13  104   88    4  381   15  297   98   32 2071   56   26\n",
            "   141    6  194    2   18    4  226   22   21  134  476   26  480    5\n",
            "   144   30    2   18   51   36   28  224   92   25  104    4  226   65\n",
            "    16   38 1334   88   12   16  283    5   16 4472  113  103   32   15\n",
            "    16    2   19  178   32]]\n",
            "tf.Tensor(\n",
            "[[[-0.03868975  0.04195425 -0.03029068 ...  0.01590535  0.0496436\n",
            "    0.02525698]\n",
            "  [-0.03868975  0.04195425 -0.03029068 ...  0.01590535  0.0496436\n",
            "    0.02525698]\n",
            "  [-0.03868975  0.04195425 -0.03029068 ...  0.01590535  0.0496436\n",
            "    0.02525698]\n",
            "  ...\n",
            "  [-0.01592961  0.00398639  0.0384778  ... -0.04412751  0.03904767\n",
            "   -0.01117608]\n",
            "  [-0.03949455 -0.0540452  -0.01487525 ...  0.04078636  0.02420986\n",
            "   -0.01701155]\n",
            "  [-0.03325516 -0.00815951 -0.02408331 ... -0.04035429  0.00969782\n",
            "   -0.0480123 ]]], shape=(1, 299, 256), dtype=float32)\n",
            "tf.Tensor(\n",
            "[[ 0.3645379   0.15277496 -0.23723035  0.35424095  0.26723638 -0.2563127\n",
            "   0.13671856  0.24727245  0.1630521  -0.18757203 -0.17929174 -0.30841517\n",
            "  -0.06295458 -0.12124313 -0.21062821  0.15632999  0.26319557 -0.35826793\n",
            "  -0.01316723 -0.26656958  0.29856452  0.3604562  -0.3006286  -0.192428\n",
            "  -0.07882854 -0.39991403  0.03139199 -0.0701409  -0.33293346 -0.16066432\n",
            "  -0.17958504 -0.23317896 -0.25580534 -0.17091101  0.30208102 -0.21083729\n",
            "   0.00692983  0.23223959 -0.29386646  0.20991391  0.20425391 -0.16724463\n",
            "   0.2530054   0.40504548  0.02074471 -0.2260246  -0.35133842 -0.39028785\n",
            "   0.19223617  0.19850998 -0.34200636  0.05103754 -0.28024843  0.34311363\n",
            "   0.44454294  0.4251098   0.282345   -0.25271592  0.23600794 -0.09396558]], shape=(1, 60), dtype=float32)\n",
            "tf.Tensor([[0.9592367]], shape=(1, 1), dtype=float32)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kNMVsBFsH95V"
      },
      "source": [
        "#### Plot learning curves"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "faKUwy8DH95V",
        "outputId": "0f8d7f53-505e-48ae-d2fb-f83269d7cacb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        }
      },
      "source": [
        "# Plot the training and validation accuracy\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "plt.style.use('ggplot')\n",
        "\n",
        "history_dict = history.history\n",
        "\n",
        "acc      = history_dict['accuracy']\n",
        "val_acc  = history_dict['val_accuracy']\n",
        "loss     = history_dict['loss']\n",
        "val_loss = history_dict['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\n",
        "\n",
        "plt.figure(figsize=(14,5))\n",
        "plt.plot(epochs, acc, marker='.', label='Training acc')\n",
        "plt.plot(epochs, val_acc, marker='.', label='Validation acc')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Classification accuracy')\n",
        "plt.legend(loc='lower right')\n",
        "plt.ylim(0, 1);"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1008x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "blC-Bzv4H95Y"
      },
      "source": [
        "#### Make predictions with the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nT0y-viOH95Y",
        "outputId": "199654b1-6180-417c-c62a-4b68236a5494",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# View the first test data example sentence\n",
        "# (invert the word index)\n",
        "inv_imdb_word_index={val:key for key,val in imdb_word_index.items()}\n",
        "[inv_imdb_word_index[index] for index in x_test[0] if index >2]\n",
        "\n"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['james',\n",
              " 'may',\n",
              " 'that',\n",
              " 'all',\n",
              " 'of',\n",
              " 'jack',\n",
              " 'in',\n",
              " 'in',\n",
              " 'a',\n",
              " 'and',\n",
              " 'use',\n",
              " 'to',\n",
              " 'and',\n",
              " 'us',\n",
              " 'playing',\n",
              " 'wrong',\n",
              " 'and',\n",
              " 'know',\n",
              " 'br',\n",
              " 'cute',\n",
              " 'cute',\n",
              " 'cute',\n",
              " 'in',\n",
              " 'in',\n",
              " 'this',\n",
              " 'made',\n",
              " 'off',\n",
              " 'him',\n",
              " 'quality',\n",
              " 'any',\n",
              " 'he',\n",
              " 'wonderfully',\n",
              " 'that',\n",
              " 'all',\n",
              " 'not',\n",
              " 'have',\n",
              " 'be',\n",
              " 'interesting',\n",
              " 'throughout',\n",
              " 'is',\n",
              " 'off',\n",
              " 'that',\n",
              " 'shows',\n",
              " 'few',\n",
              " 'is',\n",
              " '10',\n",
              " 'has',\n",
              " 'a',\n",
              " 'have',\n",
              " \"he's\",\n",
              " 'as',\n",
              " 'down',\n",
              " 'from',\n",
              " 'at',\n",
              " 'are',\n",
              " 'mother',\n",
              " 'may',\n",
              " 'that',\n",
              " 'of',\n",
              " 'jack']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hqhocrzVH95a",
        "outputId": "12012d2b-9db3-48c2-9a47-8b0a7a5ff5cc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# Get the model prediction using model.predict()\n",
        "\n",
        "model.predict(x_test[None,0,:])"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.14059912]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_5n2s8TPH95b",
        "outputId": "3a01b2e5-1fb0-4027-fbd0-3ad04d2533a3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# Get the corresponding label\n",
        "y_test[0]\n"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    }
  ]
}