Chapter6.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Chapter6.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/wllgrnt/keras-examples/blob/master/Chapter6.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6S8V9nC03dF_",
        "colab_type": "text"
      },
      "source": [
        "# Chapter 6\n",
        "## Deep Learning for text and sequences"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0LGovmMW_brs",
        "colab_type": "code",
        "outputId": "d5d4c6a4-5029-4f99-8a51-1bceca2dd9ab",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 147
        }
      },
      "source": [
        "import keras\n",
        "import numpy as np\n",
        "import string\n",
        "import zipfile\n",
        "import os\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive/')\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive/\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "omQriJnM3mIV",
        "colab_type": "text"
      },
      "source": [
        "## Working with text data\n",
        "\n",
        "Before we run a deep learning model on text, we have to convert the text into a numeric tensor. We can e.g. segment into words, segment into characters, or extract n-grams - this is tokenisation. Then we associate each token with a vector, and pack those vectors into a tensor. Here we look at two ways of associating a word with a token:\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9VcbFcZX-rGZ",
        "colab_type": "text"
      },
      "source": [
        "### One-hot encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H_2LA4Mm_ZWJ",
        "colab_type": "code",
        "outputId": "3733e195-754b-488f-d38f-6446079e14f7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 403
        }
      },
      "source": [
        "# Word-level one-hot encoding\n",
        "\n",
        "# Here we ignore special characters and punctuation.\n",
        "samples = [\"The cat sat on the mat.\", \"The dog ate my homework.\"]\n",
        "\n",
        "token_index = {}\n",
        "for sample in samples:\n",
        "  for word in sample.split():\n",
        "    if word not in token_index:\n",
        "      token_index[word] = len(token_index) + 1\n",
        "      \n",
        "# Only consider the first max_length words in each sample\n",
        "max_length = 10\n",
        "\n",
        "results = np.zeros(shape=(len(samples), max_length, max(token_index.values())+1))\n",
        "for i, sample in enumerate(samples):\n",
        "  for j, word in list(enumerate(sample.split()))[:max_length]:\n",
        "    index = token_index.get(word)\n",
        "    results[i,j,index] = 1\n",
        "    \n",
        "print(results)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[[0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]\n",
            "\n",
            " [[0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            "  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kALdA5yLAULd",
        "colab_type": "code",
        "outputId": "3b4121fe-7f61-46d8-8cb6-1375b0305b66",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 146
        }
      },
      "source": [
        "# Character-level one-hot encoding (had to modify this from the example, Challot's code doesn't run)\n",
        "\n",
        "# Here we only consider ASCII.\n",
        "samples = [\"The cat sat on the mat.\", \"The dog ate my homework.\"]\n",
        "\n",
        "characters = string.printable\n",
        "token_index = dict(zip(characters, range(1, len(characters)+1)))\n",
        "\n",
        "# Only consider the first max_length characters in each sample\n",
        "max_length = 50\n",
        "results = np.zeros(shape=(len(samples), max_length, max(token_index.values())+1))\n",
        "for i, sample in enumerate(samples):\n",
        "  for j, character in enumerate(sample):\n",
        "    index = token_index.get(character)\n",
        "    results[i,j,index] = 1\n",
        "    \n",
        "print(results[0]) # 2 samples"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " ...\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EwCOnAOfBKrB",
        "colab_type": "code",
        "outputId": "ef4fe9ad-7aaa-4ab9-8dec-b82d1b431a2f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "source": [
        "print(token_index)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'0': 1, '1': 2, '2': 3, '3': 4, '4': 5, '5': 6, '6': 7, '7': 8, '8': 9, '9': 10, 'a': 11, 'b': 12, 'c': 13, 'd': 14, 'e': 15, 'f': 16, 'g': 17, 'h': 18, 'i': 19, 'j': 20, 'k': 21, 'l': 22, 'm': 23, 'n': 24, 'o': 25, 'p': 26, 'q': 27, 'r': 28, 's': 29, 't': 30, 'u': 31, 'v': 32, 'w': 33, 'x': 34, 'y': 35, 'z': 36, 'A': 37, 'B': 38, 'C': 39, 'D': 40, 'E': 41, 'F': 42, 'G': 43, 'H': 44, 'I': 45, 'J': 46, 'K': 47, 'L': 48, 'M': 49, 'N': 50, 'O': 51, 'P': 52, 'Q': 53, 'R': 54, 'S': 55, 'T': 56, 'U': 57, 'V': 58, 'W': 59, 'X': 60, 'Y': 61, 'Z': 62, '!': 63, '\"': 64, '#': 65, '$': 66, '%': 67, '&': 68, \"'\": 69, '(': 70, ')': 71, '*': 72, '+': 73, ',': 74, '-': 75, '.': 76, '/': 77, ':': 78, ';': 79, '<': 80, '=': 81, '>': 82, '?': 83, '@': 84, '[': 85, '\\\\': 86, ']': 87, '^': 88, '_': 89, '`': 90, '{': 91, '|': 92, '}': 93, '~': 94, ' ': 95, '\\t': 96, '\\n': 97, '\\r': 98, '\\x0b': 99, '\\x0c': 100}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "URZlS-3bCnS-",
        "colab_type": "text"
      },
      "source": [
        "Keras has built-ins for both the above which take account of edge cases, special characters, and  only taking the N most common words, so we'll use those"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E3J_qOipDRgs",
        "colab_type": "code",
        "outputId": "c2e4cae4-6b7c-499f-e7b0-f23f8dc35951",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "samples = ['The cat sat on the mat.', 'The dog ate my homework.']\n",
        "tokenizer = keras.preprocessing.text.Tokenizer(num_words=1000)\n",
        "tokenizer.fit_on_texts(samples)\n",
        "\n",
        "# Turn strings into lists of integer indices\n",
        "sequences = tokenizer.texts_to_sequences(samples)\n",
        "\n",
        "# Alternatively, directly get the one-hot binary representations.\n",
        "one_hot_results = tokenizer.texts_to_matrix(samples, mode=\"binary\")\n",
        "\n",
        "word_index = tokenizer.word_index\n",
        "\n",
        "print(f\"Found {len(word_index)} unique tokens\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 9 unique tokens\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Gl3nOTncDTsL",
        "colab_type": "text"
      },
      "source": [
        "A variant of one-hot encoding is the one-hot hashing trick - rather than explicitly assigning an index to each word and keeping a reference of these indices, hash words into fixed-size vectors - this does away with the explicit word index"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Dh_MIlPyEMQ-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# One-hot with hashing\n",
        "\n",
        "samples = ['The cat sat on the mat.', 'The dog ate my homework.']\n",
        "\n",
        "# Store each word as a size-1000 vector\n",
        "dimensionality = 1000\n",
        "max_length = 10\n",
        "results=np.zeros((len(samples), max_length, dimensionality))\n",
        "\n",
        "for i, sample in enumerate(samples):\n",
        "  for j, word in list(enumerate(sample.split()))[:max_length]:\n",
        "    index = abs(hash(word)) % dimensionality # Hash to an integer\n",
        "    results[i,j,index] = 1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kRxRfoYDFFmx",
        "colab_type": "text"
      },
      "source": [
        "### Using word embeddings\n",
        "Here we use dense word vectors (as opposed to sparse, binary, high-dimensional vectors as in one-hot). These will be floating-point vectors in a low-dimensional space, and we have to learn these from the data.\n",
        "\n",
        "We can either learn the embedding jointly with the main task, or use pretrained word embeddings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fa9svqYtFHtx",
        "colab_type": "code",
        "outputId": "a2d2f420-ea53-4208-d064-bbf54237dff1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "# Learning our own embeddings, hopefully such that the geometry reflects the semantics\n",
        "\n",
        "max_features=10000\n",
        "max_length = 20\n",
        "\n",
        "# Overwrite np.load to allow keras to work with the new Numpy defaults\n",
        "old = np.load\n",
        "np.load = lambda *a,**k: old(*a,**k,allow_pickle=True)\n",
        "(X_train, y_train), (X_test, y_test) = keras.datasets.imdb.load_data(num_words=max_features)\n",
        "np.load = old\n",
        "del(old)\n",
        "\n",
        "\n",
        "\n",
        "# Turns the list of integers into a 2D integer tensor\n",
        "X_train = keras.preprocessing.sequence.pad_sequences(X_train, maxlen=max_length)\n",
        "X_test = keras.preprocessing.sequence.pad_sequences(X_test, maxlen=max_length)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://s3.amazonaws.com/text-datasets/imdb.npz\n",
            "17465344/17464789 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t9TdYKTaF65S",
        "colab_type": "code",
        "outputId": "155f373e-c35d-4814-90ed-b267a11f9667",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 533
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "# embedding has inputs: (number of possible tokens, dimensionality)\n",
        "model.add(keras.layers.Embedding(max_features, 8, input_length=max_length))\n",
        "model.add(keras.layers.Flatten())\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0618 10:40:22.205598 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "W0618 10:40:22.241738 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "W0618 10:40:22.249669 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "W0618 10:40:22.286149 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n",
            "W0618 10:40:22.303842 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3376: The name tf.log is deprecated. Please use tf.math.log instead.\n",
            "\n",
            "W0618 10:40:22.310284 140446760609664 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/nn_impl.py:180: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_1 (Embedding)      (None, 20, 8)             80000     \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 160)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 1)                 161       \n",
            "=================================================================\n",
            "Total params: 80,161\n",
            "Trainable params: 80,161\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qfOO_mamK1RU",
        "colab_type": "code",
        "outputId": "8a8be3b4-7051-4943-da4d-edb219507ebb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 459
        }
      },
      "source": [
        "history = model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0618 10:40:22.527907 140446760609664 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:986: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Train on 20000 samples, validate on 5000 samples\n",
            "Epoch 1/10\n",
            "20000/20000 [==============================] - 5s 274us/step - loss: 0.6759 - acc: 0.6050 - val_loss: 0.6398 - val_acc: 0.6814\n",
            "Epoch 2/10\n",
            "20000/20000 [==============================] - 2s 101us/step - loss: 0.5657 - acc: 0.7427 - val_loss: 0.5467 - val_acc: 0.7206\n",
            "Epoch 3/10\n",
            "20000/20000 [==============================] - 2s 101us/step - loss: 0.4752 - acc: 0.7808 - val_loss: 0.5113 - val_acc: 0.7384\n",
            "Epoch 4/10\n",
            "20000/20000 [==============================] - 2s 100us/step - loss: 0.4263 - acc: 0.8077 - val_loss: 0.5008 - val_acc: 0.7452\n",
            "Epoch 5/10\n",
            "20000/20000 [==============================] - 2s 99us/step - loss: 0.3930 - acc: 0.8258 - val_loss: 0.4981 - val_acc: 0.7538\n",
            "Epoch 6/10\n",
            "20000/20000 [==============================] - 2s 98us/step - loss: 0.3668 - acc: 0.8395 - val_loss: 0.5014 - val_acc: 0.7530\n",
            "Epoch 7/10\n",
            "20000/20000 [==============================] - 2s 100us/step - loss: 0.3435 - acc: 0.8533 - val_loss: 0.5052 - val_acc: 0.7520\n",
            "Epoch 8/10\n",
            "20000/20000 [==============================] - 2s 99us/step - loss: 0.3223 - acc: 0.8657 - val_loss: 0.5132 - val_acc: 0.7486\n",
            "Epoch 9/10\n",
            "20000/20000 [==============================] - 2s 97us/step - loss: 0.3022 - acc: 0.8766 - val_loss: 0.5213 - val_acc: 0.7490\n",
            "Epoch 10/10\n",
            "20000/20000 [==============================] - 2s 94us/step - loss: 0.2839 - acc: 0.8860 - val_loss: 0.5303 - val_acc: 0.7466\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kNJsy_nGK9SF",
        "colab_type": "text"
      },
      "source": [
        "We can get to a validation accuracy of ~75%, seems pretty good from only the first 20 words. However, flattening the embedding sequences and training a single `Dense` layer means we look at each word  separately - this is bad. We have to learn features that account for the sequence as a whole! "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KR03R-NOLgAW",
        "colab_type": "code",
        "outputId": "ff0cd9ce-038f-47cd-959c-a091970a18c2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 421
        }
      },
      "source": [
        "# Using pretrained word embeddings (could use Word2Vec, here we use GloVe)\n",
        "\n",
        "# First, download the raw text data from the IMDB dataset.\n",
        "!wget --no-check-certificate \\\n",
        "    http://mng.bz/0tIo \\\n",
        "    -O /tmp/aclimdb.zip\n",
        "zip_ref = zipfile.ZipFile(\"/tmp/aclimdb.zip\", 'r')\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-06-18 10:50:28--  http://mng.bz/0tIo\n",
            "Resolving mng.bz (mng.bz)... 35.166.24.88\n",
            "Connecting to mng.bz (mng.bz)|35.166.24.88|:80... connected.\n",
            "HTTP request sent, awaiting response... 301 Moved Permanently\n",
            "Location: https://mng.bz/0tIo [following]\n",
            "--2019-06-18 10:50:28--  https://mng.bz/0tIo\n",
            "Connecting to mng.bz (mng.bz)|35.166.24.88|:443... connected.\n",
            "WARNING: cannot verify mng.bz's certificate, issued by ‘CN=Go Daddy Secure Certificate Authority - G2,OU=http://certs.godaddy.com/repository/,O=GoDaddy.com\\\\, Inc.,L=Scottsdale,ST=Arizona,C=US’:\n",
            "  Unable to locally verify the issuer's authority.\n",
            "HTTP request sent, awaiting response... 301 \n",
            "Location: http://s3.amazonaws.com/text-datasets/aclImdb.zip [following]\n",
            "--2019-06-18 10:50:28--  http://s3.amazonaws.com/text-datasets/aclImdb.zip\n",
            "Resolving s3.amazonaws.com (s3.amazonaws.com)... 52.216.232.245\n",
            "Connecting to s3.amazonaws.com (s3.amazonaws.com)|52.216.232.245|:80... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 60711700 (58M) [application/zip]\n",
            "Saving to: ‘/tmp/aclimdb.zip’\n",
            "\n",
            "/tmp/aclimdb.zip    100%[===================>]  57.90M  50.8MB/s    in 1.1s    \n",
            "\n",
            "2019-06-18 10:50:29 (50.8 MB/s) - ‘/tmp/aclimdb.zip’ saved [60711700/60711700]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XkWypaaHNIsR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Scrape the data into lists \n",
        "imdb_dir = \"/tmp/aclImdb\"\n",
        "train_dir = os.path.join(imdb_dir, \"train\")\n",
        "\n",
        "labels = []\n",
        "texts = []\n",
        "\n",
        "for label_type in [\"neg\", \"pos\"]:\n",
        "  dir_name = os.path.join(train_dir, label_type)\n",
        "  for fname in os.listdir(dir_name):\n",
        "    if fname[-4:] == \".txt\":\n",
        "      with open(os.path.join(dir_name, fname)) as flines:\n",
        "        texts.append(flines.read())\n",
        "      labels.append(0 if label_type == \"neg\" else 1)\n",
        "      "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c2a4_VkyOd0a",
        "colab_type": "code",
        "outputId": "6d2eda2f-8608-41a2-d80c-98b819ee0356",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        }
      },
      "source": [
        "# Tokenize our text. Here, we artifically limit ourselves to just 200 training samples\n",
        "\n",
        "max_length = 20\n",
        "training_samples = 200\n",
        "validation_samples = 10000\n",
        "max_words = 10000\n",
        "\n",
        "tokenizer = keras.preprocessing.text.Tokenizer(num_words=max_words)\n",
        "tokenizer.fit_on_texts(texts)\n",
        "\n",
        "sequences = tokenizer.texts_to_sequences(texts)\n",
        "\n",
        "word_index = tokenizer.word_index\n",
        "\n",
        "print(f\"Found {len(word_index)} unique tokens\")\n",
        "\n",
        "data = keras.preprocessing.sequence.pad_sequences(sequences, maxlen=max_length)\n",
        "labels = np.asarray(labels)\n",
        "\n",
        "print(f\"Shape of data tensor: {data.shape}\")\n",
        "print(f\"Shape of label tensor: {labels.shape}\")\n",
        "\n",
        "# Shuffle and split the data\n",
        "indices = np.arange(data.shape[0])\n",
        "np.random.shuffle(indices)\n",
        "data = data[indices]\n",
        "labels = labels[indices]\n",
        "\n",
        "X_train = data[:training_samples]\n",
        "y_train = labels[:training_samples]\n",
        "X_val = data[training_samples: training_samples+validation_samples]\n",
        "y_val = labels[training_samples: training_samples+validation_samples]\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 88582 unique tokens\n",
            "Shape of data tensor: (25000, 20)\n",
            "Shape of label tensor: (25000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s_1ksv_PP8y4",
        "colab_type": "code",
        "outputId": "fab477a1-aa4a-4db1-f419-379a231501e0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 384
        }
      },
      "source": [
        "# Download the GloVe embeddings\n",
        "!wget --no-check-certificate \\\n",
        "    http://nlp.stanford.edu/data/glove.6B.zip \\\n",
        "    -O /tmp/glove.6B.zip\n",
        "zip_ref = zipfile.ZipFile(\"/tmp/glove.6B.zip\", 'r')\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-06-18 11:03:32--  http://nlp.stanford.edu/data/glove.6B.zip\n",
            "Resolving nlp.stanford.edu (nlp.stanford.edu)... 171.64.67.140\n",
            "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:80... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://nlp.stanford.edu/data/glove.6B.zip [following]\n",
            "--2019-06-18 11:03:32--  https://nlp.stanford.edu/data/glove.6B.zip\n",
            "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:443... connected.\n",
            "HTTP request sent, awaiting response... 301 Moved Permanently\n",
            "Location: http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip [following]\n",
            "--2019-06-18 11:03:32--  http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\n",
            "Resolving downloads.cs.stanford.edu (downloads.cs.stanford.edu)... 171.64.64.22\n",
            "Connecting to downloads.cs.stanford.edu (downloads.cs.stanford.edu)|171.64.64.22|:80... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 862182613 (822M) [application/zip]\n",
            "Saving to: ‘/tmp/glove.6B.zip’\n",
            "\n",
            "/tmp/glove.6B.zip   100%[===================>] 822.24M  60.5MB/s    in 13s     \n",
            "\n",
            "2019-06-18 11:03:45 (61.1 MB/s) - ‘/tmp/glove.6B.zip’ saved [862182613/862182613]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bM0eJDf8QZTH",
        "colab_type": "code",
        "outputId": "9bc11b8d-b5b0-4d8b-b33f-a5cbaa1a2101",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "# Parse the embeddings into an index mapping words to their vector reprs\n",
        "glove_path = \"/tmp/glove.6B.100d.txt\"\n",
        "\n",
        "embeddings_index = {}\n",
        "with open(glove_path) as flines:\n",
        "  for line in flines:\n",
        "    values = line.split()\n",
        "    word = values[0]\n",
        "    coefs = np.asarray(values[1:], dtype=\"float32\")\n",
        "    embeddings_index[word] = coefs\n",
        "    \n",
        "print(f\"Found {len(embeddings_index)} word vectors\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 400000 word vectors\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dAMdJzmeQ6yP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Now build an embedding matrix we can load into an Embedding layer\n",
        "\n",
        "embedding_dim = 100\n",
        "\n",
        "embedding_matrix = np.zeros((max_words, embedding_dim))\n",
        "for word, i in word_index.items():\n",
        "  if i < max_words:\n",
        "    embedding_vector = embeddings_index.get(word)\n",
        "    if embedding_vector is not None:\n",
        "      embedding_matrix[i] = embedding_vector"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qQw7AEAGRbso",
        "colab_type": "code",
        "outputId": "4ba619e3-3dad-4906-f64a-1944a980acc3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 293
        }
      },
      "source": [
        "# Define our model\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_words, embedding_dim, input_length=max_length))\n",
        "model.add(keras.layers.Flatten())\n",
        "model.add(keras.layers.Dense(32, activation=\"relu\"))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_2 (Embedding)      (None, 20, 100)           1000000   \n",
            "_________________________________________________________________\n",
            "flatten_2 (Flatten)          (None, 2000)              0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 32)                64032     \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 1)                 33        \n",
            "=================================================================\n",
            "Total params: 1,064,065\n",
            "Trainable params: 1,064,065\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BlH76XBpRoT9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Load the GloVe matrix into the Embedding layer\n",
        "model.layers[0].set_weights([embedding_matrix])\n",
        "model.layers[0].trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zDkGAJlGRywC",
        "colab_type": "code",
        "outputId": "05175292-d10e-49de-aed0-0197fa18c998",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 403
        }
      },
      "source": [
        "# Compile and train\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "history = model.fit(X_train, y_train, epochs=10, batch_size=32, validation_data=(X_val, y_val))\n",
        "model.save_weights(\"/content/drive/My Drive/Colab Notebooks/pre_trained_glove_model.h5\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 200 samples, validate on 10000 samples\n",
            "Epoch 1/10\n",
            "200/200 [==============================] - 1s 3ms/step - loss: 0.7453 - acc: 0.5450 - val_loss: 0.7165 - val_acc: 0.4969\n",
            "Epoch 2/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.5676 - acc: 0.7500 - val_loss: 0.7809 - val_acc: 0.5051\n",
            "Epoch 3/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.4935 - acc: 0.7400 - val_loss: 0.7351 - val_acc: 0.4999\n",
            "Epoch 4/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.3533 - acc: 0.9450 - val_loss: 0.9398 - val_acc: 0.5013\n",
            "Epoch 5/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.3361 - acc: 0.8600 - val_loss: 0.8728 - val_acc: 0.5067\n",
            "Epoch 6/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2509 - acc: 0.9600 - val_loss: 0.9588 - val_acc: 0.5030\n",
            "Epoch 7/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2062 - acc: 0.9750 - val_loss: 0.8176 - val_acc: 0.4989\n",
            "Epoch 8/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.1586 - acc: 0.9850 - val_loss: 1.0387 - val_acc: 0.4994\n",
            "Epoch 9/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.1360 - acc: 0.9950 - val_loss: 0.9323 - val_acc: 0.4972\n",
            "Epoch 10/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.1171 - acc: 0.9900 - val_loss: 0.8864 - val_acc: 0.4996\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tVP_0og5Sgm8",
        "colab_type": "code",
        "outputId": "792ca069-cd9b-44da-94eb-0f1ae0997a7a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "# Plot the loss and accuracy of the model over the training and validation data\n",
        "def plotLossAndAccuracy(history):\n",
        "  history_dict = history.history\n",
        "  acc = history_dict[\"acc\"]\n",
        "  val_acc = history_dict[\"val_acc\"]\n",
        "  loss = history_dict[\"loss\"]\n",
        "  val_loss = history_dict[\"val_loss\"]\n",
        "  epochs = range(1, len(loss) +1)\n",
        "\n",
        "  plt.plot(epochs, acc, \"bo\", label=\"Training Accuracy\")\n",
        "  plt.plot(epochs, val_acc, \"b\", label=\"Validation Accuracy\")\n",
        "  plt.title(\"Training and validation accuracy\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Accuracy\")\n",
        "  plt.legend()\n",
        "  plt.show()\n",
        "\n",
        "  plt.plot(epochs, loss, \"bo\", label=\"Training Loss\")\n",
        "  plt.plot(epochs, val_loss, \"b\", label=\"Validation Loss\")\n",
        "  plt.title(\"Training and validation loss\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Loss\")\n",
        "  plt.legend()\n",
        "  plt.show()\n",
        "\n",
        "plotLossAndAccuracy(history)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nE5ElC2OMMRFZsjDGGBPR/wMYDzDz6uKoHAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DWkl61jySwmm",
        "colab_type": "text"
      },
      "source": [
        "Hefty amounts of overfitting! What happens if we don't use the pretrained word embeddings, and don't freeze the embedding layer?"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "26-87rMuTMxo",
        "colab_type": "code",
        "outputId": "94de3f9e-8893-4d48-8924-8975e3307339",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 959
        }
      },
      "source": [
        "# Define our model\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_words, embedding_dim, input_length=max_length))\n",
        "model.add(keras.layers.Flatten())\n",
        "model.add(keras.layers.Dense(32, activation=\"relu\"))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "# Compile and train\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "history = model.fit(X_train, y_train, epochs=10, batch_size=32, validation_data=(X_val, y_val))\n",
        "plotLossAndAccuracy(history)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 200 samples, validate on 10000 samples\n",
            "Epoch 1/10\n",
            "200/200 [==============================] - 1s 3ms/step - loss: 0.6942 - acc: 0.4700 - val_loss: 0.6938 - val_acc: 0.4912\n",
            "Epoch 2/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.6221 - acc: 1.0000 - val_loss: 0.6948 - val_acc: 0.4978\n",
            "Epoch 3/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.5474 - acc: 1.0000 - val_loss: 0.6970 - val_acc: 0.4938\n",
            "Epoch 4/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.4386 - acc: 1.0000 - val_loss: 0.7016 - val_acc: 0.4989\n",
            "Epoch 5/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.3121 - acc: 1.0000 - val_loss: 0.7094 - val_acc: 0.4982\n",
            "Epoch 6/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2039 - acc: 1.0000 - val_loss: 0.7197 - val_acc: 0.4978\n",
            "Epoch 7/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.1280 - acc: 1.0000 - val_loss: 0.7295 - val_acc: 0.4964\n",
            "Epoch 8/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.0788 - acc: 1.0000 - val_loss: 0.7434 - val_acc: 0.4947\n",
            "Epoch 9/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.0481 - acc: 1.0000 - val_loss: 0.7553 - val_acc: 0.4916\n",
            "Epoch 10/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.0299 - acc: 1.0000 - val_loss: 0.7692 - val_acc: 0.4926\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEZGIkoKIiET+P3wy3Z49R5MAAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ea-w8e9GXeQW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Evaluate the model on the test data\n",
        "\n",
        "test_dir = os.path.join(imdb_dir, \"test\")\n",
        "\n",
        "labels = []\n",
        "texts = []\n",
        "\n",
        "for label_type in [\"neg\", \"pos\"]:\n",
        "  dir_name = os.path.join(test_dir, label_type)\n",
        "  for fname in os.listdir(dir_name):\n",
        "    if fname[-4:] == \".txt\":\n",
        "      with open(os.path.join(dir_name, fname)) as flines:\n",
        "        texts.append(flines.read())\n",
        "      labels.append(0 if label_type == \"neg\" else 1)\n",
        "      \n",
        "\n",
        "sequences = tokenizer.texts_to_sequences(texts)\n",
        "X_test = keras.preprocessing.sequence.pad_sequences(sequences, maxlen=max_length)\n",
        "y_test = np.asarray(labels)    \n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BWDatLiVX-ov",
        "colab_type": "code",
        "outputId": "c5444640-ab0b-4e2d-cf24-fbabb455f2b7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "model.load_weights(\"/content/drive/My Drive/Colab Notebooks/pre_trained_glove_model.h5\")\n",
        "model.evaluate(X_test, y_test)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "25000/25000 [==============================] - 1s 44us/step\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.8860045902633668, 0.50624]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 49
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LY9QUsX_lvSZ",
        "colab_type": "text"
      },
      "source": [
        "Only 50% accuracy!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DxubvLSsl1MQ",
        "colab_type": "text"
      },
      "source": [
        "## Recurrent Neural Networks\n",
        "\n",
        "The dense classifiers and convnets seen previously don't have any memory - each input is processed independently. So to process a sequence, you have to show the entire sequence to the network at once. These are *feedforward* networks.\n",
        "\n",
        "RNNs are different - they iterate through the sequence elemetns and maintain a state containing information relative to what they've seen so far. The state is reset between independent  elements, so one sequence is still a single data point.\n",
        "\n",
        "Pseudocode for the forward pass of a toy RNN\n",
        "\n",
        "```\n",
        "state_t = 0\n",
        "for input_t in input_sequences:\n",
        "   output_t = f(input_t, state_t)\n",
        "   state_t = output_t\n",
        "```\n",
        "\n",
        "And we know what the function `f` looks like:\n",
        "```\n",
        "f(input_t, state_t) = activation(dot(W, input_t) + dot(U, state_t) + b)\n",
        "```\n",
        "It is parameterised by two weight matrices and a bias vector."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XZqMeYB3l7Ya",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Actual, if naive, Numpy implementation of the forward pass\n",
        "\n",
        "timesteps = 100\n",
        "# Dimensionality of the input feature space\n",
        "input_features = 32\n",
        "# Dimensionality of the output feature space\n",
        "output_features = 64\n",
        "\n",
        "inputs  = np.random.random((timesteps, input_features)) # input data\n",
        "\n",
        "state_t = np.zeros((output_features,))\n",
        "\n",
        "W = np.random.random((output_features, input_features))\n",
        "U = np.random.random((output_features, output_features))\n",
        "b = np.random.random((output_features,))\n",
        "\n",
        "successive_outputs = []\n",
        "for input_t in inputs:\n",
        "  output_t = np.tanh(np.dot(W, input_t) + np.dot(U, state_t) + b)\n",
        "  \n",
        "  successive_outputs.append(output_t)\n",
        "  \n",
        "  state_t = output_t\n",
        "  \n",
        "# The final output is a 2D tensor of shape (timesteps, output_features)\n",
        "final_output_sequence = np.concatenate(successive_outputs, axis=0)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eUWWypTdpjyJ",
        "colab_type": "text"
      },
      "source": [
        "The above process is implemented in Keras as the `SimpleRNN` layer - see below for an example"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Fi1VQLhoqqyM",
        "colab_type": "code",
        "outputId": "9e2a0194-e2c1-4c9f-adcb-d9bee033da79",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(10000, 32))\n",
        "model.add(keras.layers.SimpleRNN(32))  # Only returns the last output for each sequence\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_5 (Embedding)      (None, None, 32)          320000    \n",
            "_________________________________________________________________\n",
            "simple_rnn_1 (SimpleRNN)     (None, 32)                2080      \n",
            "=================================================================\n",
            "Total params: 322,080\n",
            "Trainable params: 322,080\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nvvQLQjOqzwu",
        "colab_type": "code",
        "outputId": "2bf49b36-adf7-4480-9aab-b29129600273",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(10000, 32))\n",
        "model.add(keras.layers.SimpleRNN(32, return_sequences=True))  # Returns the full state sequence\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_7 (Embedding)      (None, None, 32)          320000    \n",
            "_________________________________________________________________\n",
            "simple_rnn_3 (SimpleRNN)     (None, None, 32)          2080      \n",
            "=================================================================\n",
            "Total params: 322,080\n",
            "Trainable params: 322,080\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hQUaWIyJrEGo",
        "colab_type": "code",
        "outputId": "01537e18-7aad-470f-e37a-7790a1dee541",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "# If you want to stack recurrent layers to increase the representational power of the network,\n",
        "# You need to have all the intermediate layers return full sequences \n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(10000, 32))\n",
        "model.add(keras.layers.SimpleRNN(32, return_sequences=True))\n",
        "model.add(keras.layers.SimpleRNN(32, return_sequences=True))\n",
        "model.add(keras.layers.SimpleRNN(32, return_sequences=True))\n",
        "model.add(keras.layers.SimpleRNN(32))\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding_8 (Embedding)      (None, None, 32)          320000    \n",
            "_________________________________________________________________\n",
            "simple_rnn_4 (SimpleRNN)     (None, None, 32)          2080      \n",
            "_________________________________________________________________\n",
            "simple_rnn_5 (SimpleRNN)     (None, None, 32)          2080      \n",
            "_________________________________________________________________\n",
            "simple_rnn_6 (SimpleRNN)     (None, None, 32)          2080      \n",
            "_________________________________________________________________\n",
            "simple_rnn_7 (SimpleRNN)     (None, 32)                2080      \n",
            "=================================================================\n",
            "Total params: 328,320\n",
            "Trainable params: 328,320\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LxIrBStirVos",
        "colab_type": "code",
        "outputId": "dabe0136-a116-4859-c435-0a12e10da094",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 156
        }
      },
      "source": [
        "# Let's use such a model on the IMDB dataset\n",
        "\n",
        "# Load the data\n",
        "\n",
        "max_features=10000\n",
        "max_length = 500\n",
        "\n",
        "print(\"Loading data...\")\n",
        "\n",
        "# Overwrite np.load to allow keras to work with the new Numpy defaults\n",
        "old = np.load\n",
        "np.load = lambda *a,**k: old(*a,**k,allow_pickle=True)\n",
        "(input_train, y_train), (input_test, y_test) = keras.datasets.imdb.load_data(num_words=max_features)\n",
        "np.load = old\n",
        "del(old)\n",
        "\n",
        "print(f\"{len(input_train)} training sequences\")\n",
        "print(f\"{len(input_test)} test sequences\")\n",
        "\n",
        "\n",
        "print(\"Pad sequences (samples x time)\")\n",
        "\n",
        "input_train = keras.preprocessing.sequence.pad_sequences(input_train, maxlen=max_length)\n",
        "input_test = keras.preprocessing.sequence.pad_sequences(input_test, maxlen=max_length)\n",
        "\n",
        "print(f\"input_train shape: {input_train.shape}\")\n",
        "print(f\"input_test shape: {input_test.shape}\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Loading data...\n",
            "Downloading data from https://s3.amazonaws.com/text-datasets/imdb.npz\n",
            "17465344/17464789 [==============================] - 1s 0us/step\n",
            "25000 training sequences\n",
            "25000 test sequences\n",
            "Pad sequences (samples x time)\n",
            "input_train shape: (25000, 500)\n",
            "input_test shape: (25000, 500)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_6daMHiPsFnw",
        "colab_type": "code",
        "outputId": "e0284586-a222-4aa4-d31e-3a0cccf473d0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 403
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_features, 32))\n",
        "model.add(keras.layers.SimpleRNN(32))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "\n",
        "history = model.fit(input_train, y_train, epochs=10, batch_size=128, validation_split=0.2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 20000 samples, validate on 5000 samples\n",
            "Epoch 1/10\n",
            "20000/20000 [==============================] - 27s 1ms/step - loss: 0.6449 - acc: 0.6173 - val_loss: 0.5848 - val_acc: 0.6854\n",
            "Epoch 2/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.4435 - acc: 0.8116 - val_loss: 0.5315 - val_acc: 0.7252\n",
            "Epoch 3/10\n",
            "20000/20000 [==============================] - 27s 1ms/step - loss: 0.3148 - acc: 0.8742 - val_loss: 0.3472 - val_acc: 0.8604\n",
            "Epoch 4/10\n",
            "20000/20000 [==============================] - 27s 1ms/step - loss: 0.2243 - acc: 0.9131 - val_loss: 0.3717 - val_acc: 0.8420\n",
            "Epoch 5/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.1457 - acc: 0.9480 - val_loss: 0.3848 - val_acc: 0.8688\n",
            "Epoch 6/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.0921 - acc: 0.9692 - val_loss: 0.5683 - val_acc: 0.8282\n",
            "Epoch 7/10\n",
            "20000/20000 [==============================] - 27s 1ms/step - loss: 0.0655 - acc: 0.9794 - val_loss: 0.5601 - val_acc: 0.7898\n",
            "Epoch 8/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.0344 - acc: 0.9903 - val_loss: 0.6178 - val_acc: 0.8138\n",
            "Epoch 9/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.0229 - acc: 0.9937 - val_loss: 0.5394 - val_acc: 0.8492\n",
            "Epoch 10/10\n",
            "20000/20000 [==============================] - 26s 1ms/step - loss: 0.0166 - acc: 0.9953 - val_loss: 0.6372 - val_acc: 0.8186\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iFb4R7ZpsrI7",
        "colab_type": "code",
        "outputId": "afdd605e-7da4-4e77-d39c-1297ab11b060",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "plotLossAndAccuracy(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nHiwAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl8VPXZ///Xxb6KCrgAQlBRFllN\ncUHcbUERqnIjFK1aLZav+1ZTl96WgvUWRcWbqlh3g9zeeqPUtVbpjyouRAQUkEIxYAAVUZFNIXL9\n/vhMhiGGMCE5cybJ+/l4zGPmnDlzzjUTmGs+u7k7IiIiAHXiDkBERLKHkoKIiCQpKYiISJKSgoiI\nJCkpiIhIkpKCiIgkKSlIlTKzuma2wczaV+WxcTKzg82syvtum9nJZlaYsr3YzPqnc+xuXOsvZnbD\n7r6+nPOONbNHq/q8Ep96cQcg8TKzDSmbTYDvgR8S2xe7e35FzufuPwDNqvrY2sDdD62K85jZRcA5\n7n58yrkvqopzS82npFDLuXvySznxS/Qid//7zo43s3ruXpyJ2EQk81R9JOVKVA/8j5k9ZWbrgXPM\n7Cgze8fMvjGz1WY20czqJ46vZ2ZuZjmJ7ScTz79sZuvN7G0z61jRYxPPDzSzf5nZOjO718zeMrPz\ndxJ3OjFebGZLzexrM5uY8tq6ZnaXma01s2XAgHI+nxvNbGqpfZPMbELi8UVmtijxfv6d+BW/s3MV\nmdnxicdNzOyJRGwLgMNLHXuTmS1LnHeBmQ1O7O8O/DfQP1E192XKZ3tLyut/k3jva83sOTPbP53P\nZlfM7IxEPN+Y2RtmdmjKczeY2Soz+9bMPk55r0ea2ZzE/s/NbHy615MIuLtuuuHuAIXAyaX2jQW2\nAKcTfkQ0Bn4CHEEoaR4I/Au4NHF8PcCBnMT2k8CXQC5QH/gf4MndOHYfYD0wJPHc1cBW4PydvJd0\nYnweaAHkAF+VvHfgUmAB0A5oCcwM/1XKvM6BwAagacq5vwByE9unJ44x4ERgM9Aj8dzJQGHKuYqA\n4xOP7wD+AewFdAAWljp2GLB/4m/yi0QM+yaeuwj4R6k4nwRuSTz+aSLGXkAj4M/AG+l8NmW8/7HA\no4nHXRJxnJj4G90ALE487gYsB/ZLHNsRODDxeDYwIvG4OXBE3P8XavNNJQVJx5vu/ld33+bum919\ntru/6+7F7r4MmAwcV87rn3H3AnffCuQTvowqeuwgYK67P5947i5CAilTmjH+yd3XuXsh4Qu45FrD\ngLvcvcjd1wK3lXOdZcBHhGQFcArwtbsXJJ7/q7sv8+AN4HWgzMbkUoYBY939a3dfTvj1n3rdp919\ndeJvMoWQ0HPTOC/ASOAv7j7X3b8D8oDjzKxdyjE7+2zKMxyY7u5vJP5GtxESyxFAMSEBdUtUQX6S\n+OwgJPdOZtbS3de7+7tpvg+JgJKCpOPT1A0z62xmL5rZZ2b2LTAGaFXO6z9LebyJ8huXd3Zsm9Q4\n3N0Jv6zLlGaMaV2L8Au3PFOAEYnHv0hsl8QxyMzeNbOvzOwbwq/08j6rEvuXF4OZnW9m8xLVNN8A\nndM8L4T3lzyfu38LfA20TTmmIn+znZ13G+Fv1NbdFwPXEP4OXySqI/dLHHoB0BVYbGbvmdmpab4P\niYCSgqSjdHfMBwi/jg929z2A3xOqR6K0mlCdA4CZGTt+iZVWmRhXAwekbO+qy+zTwMlm1pZQYpiS\niLEx8AzwJ0LVzp7A39KM47OdxWBmBwL3AaOBlonzfpxy3l11n11FqJIqOV9zQjXVyjTiqsh56xD+\nZisB3P1Jd+9HqDqqS/hccPfF7j6cUEV4J/CsmTWqZCyym5QUZHc0B9YBG82sC3BxBq75AtDHzE43\ns3rAFUDriGJ8GrjSzNqaWUvg+vIOdvfPgDeBR4HF7r4k8VRDoAGwBvjBzAYBJ1UghhvMbE8L4zgu\nTXmuGeGLfw0hP/6aUFIo8TnQrqRhvQxPAReaWQ8za0j4cv6nu++05FWBmAeb2fGJa19HaAd618y6\nmNkJiettTty2Ed7AuWbWKlGyWJd4b9sqGYvsJiUF2R3XAOcR/sM/QGgQjpS7fw6cDUwA1gIHAR8Q\nxlVUdYz3Eer+PyQ0gj6TxmumEBqOk1VH7v4NcBUwjdBYO5SQ3NLxn4QSSyHwMvB4ynnnA/cC7yWO\nORRIrYd/DVgCfG5mqdVAJa9/hVCNMy3x+vaEdoZKcfcFhM/8PkLCGgAMTrQvNARuJ7QDfUYomdyY\neOmpwCILvdvuAM529y2VjUd2j4WqWZHqxczqEqorhrr7P+OOR6SmUElBqg0zG5CoTmkI3EzotfJe\nzGGJ1ChKClKdHAMsI1RN/Aw4w913Vn0kIrtB1UciIpKkkoKIiCRVuwnxWrVq5Tk5OXGHISJSrbz/\n/vtfunt53biBapgUcnJyKCgoiDsMEZFqxcx2NTIfUPWRiIikUFIQEZEkJQUREUlSUhARkSQlBRER\nSaoVSSE/H3JyoE6dcJ9foaXoRURqj2rXJbWi8vNh1CjYtClsL18etgFGVnpeSBGRmqXGlxRuvHF7\nQiixaVPYLyIiO6rxSWHFiortFxGpzWp8Umi/k4UUd7ZfRCQb/fvfkIn5S2t8Uhg3Dpo02XFfkyZh\nv4hIdfDii9C9O9x7b/TXqvFJYeRImDwZOnQAs3A/ebIamUWkenjsMRgyBLp2heHDo79eje99BCEB\nKAmISHUzfjz89rdw8snwf/8HzZtHf80aX1IQEalutm2Da68NCeHss+GFFzKTEKCWlBRERKqLrVvh\nwgvhiSfgssvg7rvDwNtMUVIQEckSGzfCf/wHvPwyjB0LN9wQ2kIzSdVHIpJV3OFvf4O+feHoo2Hp\n0rgjyoy1a+Gkk+DVV0NnmBtvzHxCACUFEckis2eHRtWf/Qy++AIWLYI+fWr+fGUrVsAxx8DcufDs\ns/DrX8cXi5KCiMRuyRIYNiyUDubNC/XoixeHxz17wjnnwPnnw4YNcUda9RYuhH79YPXqUEL6+c/j\njSfSpGBmA8xssZktNbO8nRwzzMwWmtkCM5sSZTwikl1Wr4bRo6FLlzBA6+abYdkyuOIKaNgwzDww\nYwb8/veh4fXww+GDD+KOuuq8/XYoIRQXw8yZcOyxcUcUYVIws7rAJGAg0BUYYWZdSx3TCfgd0M/d\nuwFXRhWPiGSPdevgppvg4IPhL3+Biy8O0ziMGQN77LHjsfXqwR/+AG+8ERpijzwylCQyMeVDlF58\nMbQhtGoFs2ZBjx5xRxREWVLoCyx192XuvgWYCgwpdcyvgUnu/jWAu38RYTwiErPvv4e77oKDDgpT\nzZx+emg3mDQJ9tuv/Nced1yoThowAK66Krx2zZrMxF3VUkcpv/kmdOwYd0TbRZkU2gKfpmwXJfal\nOgQ4xMzeMrN3zGxAWScys1FmVmBmBWuq678CkVrshx/g8cfhkEPg6quhd28oKICpU0NpIV0tW8Jz\nz4U5gP7+99De8MYb0cUdhfHjQ/vICSeEqrF99ok7oh3F3dBcD+gEHA+MAB40sz1LH+Tuk909191z\nW7duneEQRWR3uYdqkl694LzzQlXJ3/4Gr70W2gd2hxlceim8+y60aBF6K914Yxj0lc3iHKVcEVEm\nhZXAASnb7RL7UhUB0919q7t/AvyLkCREpJp75x04/ngYNAg2bw6lgtmz4ZRTqub8PXuG0savfgW3\n3hqqlwoLq+bcVW3r1pAU77wzjFKeMiU0pGejKJPCbKCTmXU0swbAcGB6qWOeI5QSMLNWhOqkZRHG\nJCIRW7QIzjgDjjoqdCudNCl0uzz77KqfrqFp09BQPXUqLFgQSiT/+79Ve43K2rgxtB88+WQYpXzP\nPZmdtqKiIgvN3YuBS4FXgUXA0+6+wMzGmNngxGGvAmvNbCEwA7jO3ddGEc+778Lll8PK0mUVEakS\nRUVw0UVw2GHw+uuhJ9HSpfD//h80aBDttc8+Owz86tw5jHdIXZc9TtkySrlC3L1a3Q4//HDfHRMn\nuter596ggfsll7ivWLFbpxGRUr76yv23v3Vv1Mi9fn33K65w/+KLeGLZssU9L8/dzL1LF/f58+OJ\nw919+XL3zp3dGzZ0nzYtvjhKAAWexndsFhdiqtZll8G//hXq9R54IPR4GD1aazWL7K7Nm+H22+HA\nA0OPmqFDQ3XR3XdDXP1B6teHP/0pNGZ//TX85Cfw5z9nfkzDggXZNUq5ImpNUoDQF3jy5FCkveAC\neOihkBwuvjh7G6hEsk1xcfi/06kTXH99aDv44IMw4jhb+tuffHIY03DiiXDJJXDWWfDVV5m59qxZ\n0L9/do1SrohalRRKdOgA998fksNFF8Gjj4Z/4L/+dRhiLyI/5h7GCPToEf7ftG0b+tm/9FLoCZRt\n9tkndPu8885w36sX/POf0V7zxRdDQsq2UcoVUSuTQon27UPR8t//ht/8JvzSOeSQ0MXt3/+OOzqp\nbTZtCrOBTp8Ob70FH38MX34ZBn7FbebMUB1yxhmhv/2zz27vcprN6tQJg+Xefjt0AT3++NAAHsVn\nms2jlCvCvJpNIJKbm+sFBQWRnHvlylBH+sADoeh3zjmht0AnjZyQDLjiCpg48cf7zWCvvcJo3pYt\nw6/Q0o/L2le/fuVj+vBD+N3vwi/gNm3glltC1Wu9arg81/r1oSrpiSdClU5+PrRrVzXnjmMt5Yoy\ns/fdPXeXxykp/Njq1SE53H8/bNkCv/hFmLzr0EMjvazUYh99tH3U7+jRoSvj2rWhpFDe4/K6XTZv\nXn7SKGtf48bhtcuXb5+ZdI89IC8vdOlu0iQzn0eUnnhiezfZhx8Ov+5317ZtIRnceWfoFvvYY9k7\nKE1JoQp89ln4BXDffWEir+HDQ3Lo0iUjl5dawj38wvzgg7CuQMuW6b928+byk0ZZ+779dufna9w4\nJIjPPw8llMsuCyWFvfeu/PvMJkuWwIgR8P77YcqM8eOhUaOKnWPr1lDV/OST8aylXFFKClXoiy/g\njjvCyMzNm8MvgptvDnWHIpX17LOhO+ekSeEXbNS2bAk9ccorieyxR5in54ADdn2+6mrLlpDwJkwI\nDcJTp6b/gy8b1lKuKCWFCKxZE4qJ//3fodg+dGgoYh92WCzhSA2waVP4Itpzz/CrtTrW1Vd3L78c\nqu02bgxtOr/6Vflf8GvXwmmnhXmc7r8/3qUzKyLdpJDFhZ3s07o13HZbGNOQlxf+MXXvHpLD/Plx\nRyfV0fjxYQDlxIlKCHEZODCMaTj66NDVdsSIsAhQWbJpLeWoKCnshlatwqyMy5eHNobXXgv9tM88\nM/xjEUnH8uXhR8bZZ4cZPiU+++8f5ie67bbwZd+rV+hym6o6j1KuCCWFSth7b/jjH0PJ4fe/D4t9\n9O4dejPMmRN3dJLtrr02VFOMHx93JAKhkfj667cPcDvmmJAktm2r/qOUK0JJoQrstVdYQ7awMPTj\nnjkzLCBy+ulhvneR0t54A555JjRQ1uTG3OroyCNDiX/o0NAQ3b9/9R+lXBFqaI7AunVhucAJE8Kk\nXKeeCv/5n9C3b9yRSTYoLg7VE5s2hXUGKtoVUjLDHR55JHQ37dIlTOeRbUtnVoQammPUokVoaygs\nDIuTv/MOHHFEWHD87bfjjk7idt99oX56wgQlhGxmFnoirVgRSgjVOSFUhJJChPbYI1QPFBaG6XwL\nCkIPh1/+MvNT+Up2WLMmtD+dckrlRtJK5rRsGf0iQdlESSEDmjcPXVgLC+Gqq8Iw++efjzsqicNN\nN8GGDWFJxmwf7CS1k5JCBjVrFuZUOuywMHPjd9/FHZFk0pw58OCD2+uoRbKRkkKG1asXfiV+8kkY\nHS21g3tIBq1bh04HItlKSSEGJ54YVoK69daw2LnUfFOmhMbKP/0pdEQQyVZKCjG5444wKOb66+OO\nRKK2fj1cdx3k5sL558cdjUj5lBRikpMTviimTAmrNEnNdeutYWqEe+/N7qmVRUBJIVbXXx9Wfrr8\n8uxYclGq3pIlYTzCeeeFkbIi2S7SpGBmA8xssZktNbO8Mp4/38zWmNncxO2iKOPJNk2bhmqkDz4I\nK0BJzXP11WElrj/9Ke5IRNITWVIws7rAJGAg0BUYYWZlLUvzP+7eK3H7S1TxZKthw8LkWjfcEKbE\nkJrjpZfghRfCYLX99487GpH0RFlS6Assdfdl7r4FmApoDGcpZqGL6ldfhUn1pGbYsgWuvBIOOSRU\nD4pUF1EmhbbApynbRYl9pZ1lZvPN7BkzK3O+SDMbZWYFZlawZs2aKGKNVa9eMGpUWNFt4cK4o5Gq\ncM89oT3hnntq1xQJUv3F3dD8VyDH3XsArwGPlXWQu09291x3z23dunVGA8yUP/4xTIdxxRWaF6m6\nW70axowJU6cPGBB3NCIVE2VSWAmk/vJvl9iX5O5r3f37xOZfgMMjjCertWoVvkj+/nfNi1Td5eWF\n6qMJE+KORKTiokwKs4FOZtbRzBoAw4HpqQeYWWrz22BgUYTxZL3Ro6FbN82LVJ29/TY8/jhccw0c\nfHDc0YhUXGRJwd2LgUuBVwlf9k+7+wIzG2NmgxOHXW5mC8xsHnA5cH5U8VQHqfMi6Vdm9bNtW5jf\nqE2b0JtMpDrSymtZ6Kyz4JVXYPHiMLhNqoeHHoKLLoL8fPjFL+KORmRHWnmtGrvjjjDCWfMiVR/f\nfBPW8+3XD0aMiDsakd2npJCFOnaE3/42zIv01ltxRyPp+MMf4Msvw/xGWjxHqjMlhSxVMi/SZZdp\nXqRst3BhSAajRkHv3nFHI1I5SgpZqmlTGD9e8yJlO/cwYrl5cxg7Nu5oRCpPSSGLnX029O8ferJ8\n803c0UhZnnsOXn89DD5s1SruaEQqT0khi5nBxIlhXqRbbok7Gilt8+YwpuSww+A3v4k7GpGqoaSQ\n5Xr1gl//WvMiZaM77oDCwpC469WLOxqRqqGkUA2MHat5kbLNihVhjYT/+A844YS4oxGpOkoK1UCr\nVqHLo+ZFyh7XXRfux4+PNw6RqqakkEH5+WFt5jp1wn1+fvqv1bxI2eMf/4Cnnw4T33XoEHc0IlVL\nSSFD8vNDP/bly0MV0PLlYTvdxFC/vuZFygbFxaELaocO20sLIjWJkkKG3HgjbNq0475Nm8L+dJ10\nEpxxBowbB0VFVRufpOeBB+DDD0Nibtw47mhEqp6SQoasWFGx/Ttz552aFykuX34JN9+8PTmL1ERK\nChnSvn3F9u9Mx46h2kLzImXezTfDt9+GajzNbyQ1lZJChowbB02a7LivSZOwv6Ly8sK8SJdfrnmR\nMuWDD0LV0aWXhgZ/kZpKSSFDRo6EyZNDA6VZuJ88OeyvqKZN4fbbYc4czYuUCSXzG7VsqZHlUvNp\nkZ1qyh2OPRY+/hiWLIE994w7oprrqafCojkPPhgW0RGpjrTITg1XMi/S2rVhYJtEY8MGuPZaOPxw\nuOCCuKMRiZ6SQjXWu3eYF+neezUvUlRuvRVWrQoJuG7duKMRiZ6SQjVXMi/SlVdqXqSqtnRp6AJ8\n7rlw9NFxRyOSGUoK1Vzr1qH66LXXYPr0uKOpWa6+Gho0gNtuizsSkcxRUqgBRo+Grl3hqqs0L1JV\nefll+Otf4aaboE2buKMRyZxIk4KZDTCzxWa21MzyyjnuLDNzM9tly7j8mOZFqlpbtoTquIMPDvci\ntUlkScHM6gKTgIFAV2CEmXUt47jmwBXAu1HFUhucfLLmRaoqEyfCv/4Fd98NDRvGHY1IZkVZUugL\nLHX3Ze6+BZgKDCnjuD8C/wWo4qOSNC9S5a1eHdpoTj0VTjst7mhEMi/KpNAW+DRluyixL8nM+gAH\nuPuL5Z3IzEaZWYGZFaxZs6bqI60hOnYMfeo1L9Lu+93v4Pvv4a674o5EJB6xNTSbWR1gAnDNro51\n98nunuvuua1bt44+uGrsd7+Dtm01L9LueOcdeOyx0GB/yCFxRyMSjyiTwkrggJTtdol9JZoDhwH/\nMLNC4EhguhqbK6dp07BE5Jw58MgjcUdTfWzbBpddBvvtF3ocidRWUSaF2UAnM+toZg2A4UCyJ727\nr3P3Vu6e4+45wDvAYHfXxEaVNHw4HHNMKDV8803c0VQPjz4KBQVhosHmzeOORiQ+9aI6sbsXm9ml\nwKtAXeBhd19gZmOAAnfXUKuIlMyLdPjhodFU9ePbuYdEuXz5jrfHHoOjjtq9WWtFahLNklqDXXxx\nmFp73rwwuK02cIfPP//xl37JrbAQ1q/f8TWNGsGhh8Ljj0OPHrGELRK5dGdJVVKowdasgU6doG9f\nePXVmrFaWHExrFy58y/95ctD76FULVqE9Ss6dICcnO2PS26tW9eMz0akPOkmhciqjyR+JfMiXXll\nmBdpSFmjRLLMd9+Fdat39oVfVPTjXlX77BO+3Hv2hMGDf/yl36JFPO9FpDpSSaGG27oVevUKX7YL\nFoSqkmywdi28/Xa4LV26/Uv/s892PK5OndDFtqxf+B06hDWuGzeO5S2IVCsqKQiwfV6kU04J8yLd\ncEPmY3CHxYvDgLpZs8L94sXhuXr1wqC7Dh3CCOLSX/pt24b3ICKZoZJCLXHGGWF67cWLwxdtlDZt\nCt07S5LArFnw1Vfhub33DmsTHH009OsHubnQpEm08YiISgpSyp13hh5I118PTz5ZtedetWrHBDBn\nTmgQBujcGX7+85AAjj469PJRo65I9lJSqCUOPDDMizRuXFh/oV+/3TvPDz/Ahx/uWBW0fHl4rlGj\n0NPp2mvD+Y88Elq1qrr3ICLRU/VRLbJxY/ilvu++8N576a05vG4dvPvu9iTwzjthMXuA/fffXgLo\n1y80aDdoEO17EJHdU6XVR2Z2EFDk7t+b2fFAD+Bxd9ckCtVI06ZhGoeRI8O8SBddtOPz7mGhnpIS\nwKxZoVTgHnoBde8Ov/zl9iTQoYOqgkRqmrRKCmY2F8gFcoCXgOeBbu5+aqTRlUElhcpxh/79wyIy\nH30Ey5btmARKuoQ2bx6mfShJAH37wh57xBu7iOy+qm5o3paYy+gM4F53v9fMPqhciBKHknmRcnPD\njKAlvwk6dgyrt5UkgW7d0qteEpGaJd2ksNXMRgDnAacn9qn3eDXVpw9MmhQGjZV0D91//7ijEpFs\nkG5SuAD4DTDO3T8xs47AE9GFJVEbPTruCEQkG6WVFNx9IXA5gJntBTR39/+KMjAREcm8tBbZMbN/\nmNkeZrY3MAd40MwmRBuaiIhkWrorr7Vw92+BMwldUY8ATo4uLBERiUO6SaGeme0PDANeiDAeERGJ\nUbpJYQxhWc1/u/tsMzsQWBJdWCIiEoe0koK7/6+793D30YntZe5+VrShSVTy88P6BHXqhPv8/Lgj\nEpFskW5Dczszm2ZmXyRuz5pZu6iDk6qXnw+jRoVJ7NzD/ahRSgwiEqRbffQIMB1ok7j9NbFPqpkb\nbwzrHaTatCnsFxFJNym0dvdH3L04cXsUaB1hXBKRFSsqtl9Eapd0k8JaMzvHzOombucAa6MMTKLR\nvn3F9otI7ZJuUvgVoTvqZ8BqYChw/q5eZGYDzGyxmS01s7wynv+NmX1oZnPN7E0z61qB2GU3jBv3\n4+UvmzQJ+0VE0u19tNzdB7t7a3ffx91/DpTb+8jM6gKTgIFAV2BEGV/6U9y9u7v3Am4HNEo6YiNH\nwuTJ29dC6NAhbI8cGXdkIpIN0i0plOXqXTzfF1ia6L66BZgKDEk9IDFKukRToHotA1dNjRwJhYWw\nbVu4V0IQkRKVWaN5V2tutQU+TdkuAo740UnMLiEkmAbAiWVeyGwUMAqgvSq/RUQiU5mSQpX8qnf3\nSe5+EHA9cNNOjpns7rnuntu6tTo9iYhEpdySgpmtp+wvfwMa7+LcK4EDUrbbJfbtzFTgvl2cU0RE\nIlRuUnD35pU492ygU2JBnpXAcOAXqQeYWSd3L5lD6TQ0n5KISKwq06ZQrsSazpcSJtKrCzzs7gvM\nbAxQ4O7TgUvN7GRgK/A1YblPERGJSWRJAcDdXwJeKrXv9ymPr4jy+iIiUjGVaWgWEZEaRklBRESS\nlBRERCRJSUFERJKUFEREJElJQUREkpQUREQkSUlBRESSlBRERCRJSUFERJKUFEREJElJQUREkpQU\nREQkSUlBRESSlBRERCRJSUFik58POTlQp064z8+POyIRiXSRHZGdyc+HUaNg06awvXx52AYYOTK+\nuERqO5UUJBY33rg9IZTYtCnsF5H4KClILFasqNh+EckMJQWJRfv2FdsvIpmhpCCxGDcOmjTZcV+T\nJmG/iMRHSUFiMXIkTJ4MHTqAWbifPFmNzCJxizQpmNkAM1tsZkvNLK+M5682s4VmNt/MXjezDlHG\nI9ll5EgoLIRt28K9EoJI/CJLCmZWF5gEDAS6AiPMrGupwz4Act29B/AMcHtU8YiIyK5FWVLoCyx1\n92XuvgWYCgxJPcDdZ7h7ScfEd4B2EcYjIiK7EGVSaAt8mrJdlNi3MxcCL5f1hJmNMrMCMytYs2ZN\nFYYoIiKpsqKh2czOAXKB8WU97+6T3T3X3XNbt26d2eBERGqRKKe5WAkckLLdLrFvB2Z2MnAjcJy7\nfx9hPCIisgtRlhRmA53MrKOZNQCGA9NTDzCz3sADwGB3/yLCWEREJA2RJQV3LwYuBV4FFgFPu/sC\nMxtjZoMTh40HmgH/a2ZzzWz6Tk4nIiIZEOksqe7+EvBSqX2/T3l8cpTXFxGRismKhmYREckOSgoi\nIpKkpCAiIklKCiIikqSkICIiSUoKIiKSpKQgIiJJSgoiIpKkpCC1Xn4+5ORAnTrhPj8/7ohE4hPp\niGaRbJefD6NGwabEqh7Ll4dt0EpwUjuppCC12o03bk8IJTZtCvtFaiMlBanVVqyo2H6Rmk5JQWq1\n9u0rtl+kplNSkFpt3Dho0mTHfU2ahP0itZGSgtRqI0fC5MnQoQOYhfvJk9XILLWXeh9JrTdypJKA\nSAmVFEREJElJQUREkpQUREQkSUlBRESSlBRERCRJSUFERJKUFEREJCnSpGBmA8xssZktNbO8Mp4/\n1szmmFmxmQ2NMhYREdm1yAavmVldYBJwClAEzDaz6e6+MOWwFcD5wLWVudbWrVspKiriu+++q8xp\nJAaNGjWiXbt21K9fP+5QRIRTOx+AAAAQWUlEQVRoRzT3BZa6+zIAM5sKDAGSScHdCxPPbavMhYqK\nimjevDk5OTmYWWVOJRnk7qxdu5aioiI6duwYdzixys8P03WvWBEm4xs3TqOsJR5RVh+1BT5N2S5K\n7KswMxtlZgVmVrBmzZofPf/dd9/RsmVLJYRqxsxo2bJlrS/hlSz0s3w5uG9f6EcrwEkcqkVDs7tP\ndvdcd89t3bp1mccoIVRP+rtpoR/JLlEmhZXAASnb7RL7RCSFFvqRbBJlUpgNdDKzjmbWABgOTI/w\nemmr6oXa165dS69evejVqxf77bcfbdu2TW5v2bIlrXNccMEFLF68uNxjJk2aRH4V1Skcc8wxzJ07\nt0rOJZWjhX4km0TW0OzuxWZ2KfAqUBd42N0XmNkYoMDdp5vZT4BpwF7A6Wb2B3fvFlVMEM1C7S1b\ntkx+wd5yyy00a9aMa6/dsUOVu+Pu1KlTdh5+5JFHdnmdSy65ZPcClKw2btyO/yZBC/1IfCJtU3D3\nl9z9EHc/yN3HJfb93t2nJx7Pdvd27t7U3VtGnRAgs/W3S5cupWvXrowcOZJu3bqxevVqRo0aRW5u\nLt26dWPMmDHJY0t+uRcXF7PnnnuSl5dHz549Oeqoo/jiiy8AuOmmm7j77ruTx+fl5dG3b18OPfRQ\nZs2aBcDGjRs566yz6Nq1K0OHDiU3NzftEsHmzZs577zz6N69O3369GHmzJkAfPjhh/zkJz+hV69e\n9OjRg2XLlrF+/XoGDhxIz549Oeyww3jmmWeq8qOrVbTQj2STatHQXJUyXX/78ccfc9VVV7Fw4ULa\ntm3LbbfdRkFBAfPmzeO1115j4cKFP3rNunXrOO6445g3bx5HHXUUDz/8cJnndnfee+89xo8fn0ww\n9957L/vttx8LFy7k5ptv5oMPPkg71okTJ9KwYUM+/PBDnnjiCc4991y2bNnCn//8Z6699lrmzp3L\n7NmzadOmDS+99BI5OTnMmzePjz76iFNOOWX3PiABQgIoLIRt28K9EoLEpdYlhUzX3x500EHk5uYm\nt5966in69OlDnz59WLRoUZlJoXHjxgwcOBCAww8/nMLCwjLPfeaZZ/7omDfffJPhw4cD0LNnT7p1\nS7/w9eabb3LOOecA0K1bN9q0acPSpUs5+uijGTt2LLfffjuffvopjRo1okePHrzyyivk5eXx1ltv\n0aJFi7SvIyLZq9YlhUwv1N60adPk4yVLlnDPPffwxhtvMH/+fAYMGFBmH/0GDRokH9etW5fi4uIy\nz92wYcNdHlMVzj33XKZNm0bDhg0ZMGAAM2fOpEuXLhQUFNCtWzfy8vK49dZbI7u+ZE5Vd8KQ6qfW\nJYU462+//fZbmjdvzh577MHq1at59dVXq/wa/fr14+mnnwZCW0BZJZGd6d+/f7J306JFi1i9ejUH\nH3wwy5Yt4+CDD+aKK65g0KBBzJ8/n5UrV9KsWTPOPfdcrrnmGubMmVPl70UyS4PoBKKd5iJrxbVQ\ne58+fejatSudO3emQ4cO9OvXr8qvcdlll/HLX/6Srl27Jm87q9r52c9+lpxzqH///jz88MNcfPHF\ndO/enfr16/P444/ToEEDpkyZwlNPPUX9+vVp06YNt9xyC7NmzSIvL486derQoEED7r///ip/L5JZ\n5XXCUBtH7WHuHncMFZKbm+sFBQU77Fu0aBFdunSJKaLsUlxcTHFxMY0aNWLJkiX89Kc/ZcmSJdSr\nl735X3+/7FCnTighlGYWGsClejOz9909d1fHZe83heyWDRs2cNJJJ1FcXIy788ADD2R1QpDs0b59\nqDIqa7/UHvq2qGH23HNP3n///bjDkGpIg+gEamFDs4iUTYPoBFRSEJEUcXXCkOyhkoKIiCQpKYhI\n1tEguvgoKVSBE0444UcD0e6++25Gjx5d7uuaNWsGwKpVqxg6dGiZxxx//PGU7oJb2t13382mlNbB\nU089lW+++Sad0Mt1yy23cMcdd1T6PCIVoUF08VJSqAIjRoxg6tSpO+ybOnUqI0aMSOv1bdq0qdQs\no6WTwksvvcSee+652+cTiZNWootXjWtovvJKqOq1Y3r1gsSM1WUaOnQoN910E1u2bKFBgwYUFhay\natUq+vfvz4YNGxgyZAhff/01W7duZezYsQwZMmSH1xcWFjJo0CA++ugjNm/ezAUXXMC8efPo3Lkz\nmzdvTh43evRoZs+ezebNmxk6dCh/+MMfmDhxIqtWreKEE06gVatWzJgxg5ycHAoKCmjVqhUTJkxI\nzrJ60UUXceWVV1JYWMjAgQM55phjmDVrFm3btuX555+ncePGaX0eZZ1z48aNDBs2jKKiIn744Qdu\nvvlmzj77bPLy8pg+fTr16tXjpz/9qUoesktaiS5eNS4pxGHvvfemb9++vPzyywwZMoSpU6cybNgw\nzIxGjRoxbdo09thjD7788kuOPPJIBg8evNO1ie+77z6aNGnCokWLmD9/Pn369Ek+N27cOPbee29+\n+OEHTjrpJObPn8/ll1/OhAkTmDFjBq1atdrhXO+//z6PPPII7777Lu7OEUccwXHHHcdee+3FkiVL\neOqpp3jwwQcZNmwYzz77bHKG1PLs7JzLli2jTZs2vPjii0CY/nvt2rVMmzaNjz/+GDOrkiotqfmy\naRBdfn4ooaxYEa4/blzN751V45JCeb/oo1RShVSSFB566CEgrHlwww03MHPmTOrUqcPKlSv5/PPP\n2W+//co8z8yZM7n88ssB6NGjBz169Eg+9/TTTzN58mSKi4tZvXo1Cxcu3OH50t58803OOOOM5Eyt\nZ555Jv/85z8ZPHgwHTt2pFevXkD503One84BAwZwzTXXcP311zNo0CD69++fnG7jwgsvZNCgQQwa\nNCita0jtli2D6KJYpbE6UJtCFRkyZAivv/46c+bMYdOmTRx++OEA5Ofns2bNGt5//33mzp3Lvvvu\nW+Z02bvyySefcMcdd/D6668zf/58TjvttN06T4mSabehaqbePuSQQ5gzZw7du3fnpptuYsyYMdSr\nV4/33nuPoUOH8sILLzBgwIBKXUNqh2wZRJdNbRuZ7I2lpFBFmjVrxgknnMCvfvWrHRqY161bxz77\n7EP9+vWZMWMGy8sqF6c49thjmTJlCgAfffQR8+fPB8K0202bNqVFixZ8/vnnvPzyy8nXNG/enPXr\n1//oXP379+e5555j06ZNbNy4kWnTptG/f/9Kvc+dnXPVqlU0adKEc845h+uuu445c+awYcMG1q1b\nx6mnnspdd93FvHnzKnVtqT2yYSW6bGnbyHRvrBpXfRSnESNGcMYZZ+zQE2nkyJGcfvrpdO/endzc\nXDp37lzuOUaPHs0FF1xAly5d6NKlS7LE0bNnT3r37k3nzp054IADdph2e9SoUQwYMIA2bdowY8aM\n5P4+ffpw/vnn07dvXyA0Cvfu3TvtqiKAsWPHJteFBigqKirznK+++irXXXcdderUoX79+tx3332s\nX7+eIUOG8N133+HuTJgwIe3risQtW9o2Mj2luabOltjp7yfZqHSbAoS2jUxXZVXVlObpTp2t6iMR\nkTJkS9tGpteVjzQpmNkAM1tsZkvNLK+M5xua2f8knn/XzHKijEdEpCKyoW0j0+vKR5YUzKwuMAkY\nCHQFRphZ11KHXQh87e4HA3cB/7W716tu1WAS6O8mUr5Ml1iiLCn0BZa6+zJ33wJMBYaUOmYI8Fji\n8TPASbazUV3laNSoEWvXrtUXTDXj7qxdu5ZGjRrFHYpIVstkiSXK3kdtgU9TtouAI3Z2jLsXm9k6\noCXwZepBZjYKGAXQvoyKtHbt2lFUVMSaNWuqLHjJjEaNGtGuXbu4wxCRhGrRJdXdJwOTIfQ+Kv18\n/fr16dixY8bjEhGpaaKsPloJHJCy3S6xr8xjzKwe0AJYG2FMIiJSjiiTwmygk5l1NLMGwHBgeqlj\npgPnJR4PBd5wNQyIiMQmsuqjRBvBpcCrQF3gYXdfYGZjgAJ3nw48BDxhZkuBrwiJQ0REYlLtRjSb\n2Rqg/AmEsl8rSjWm13L6PLbTZ7EjfR47qszn0cHdW+/qoGqXFGoCMytIZ7h5baHPYzt9FjvS57Gj\nTHwemuZCRESSlBRERCRJSSEek+MOIMvo89hOn8WO9HnsKPLPQ20KIiKSpJKCiIgkKSmIiEiSkkIG\nmdkBZjbDzBaa2QIzuyLumOJmZnXN7AMzeyHuWOJmZnua2TNm9rGZLTKzo+KOKU5mdlXi/8lHZvaU\nmdWa6XTN7GEz+8LMPkrZt7eZvWZmSxL3e0VxbSWFzCoGrnH3rsCRwCVlrDFR21wBLIo7iCxxD/CK\nu3cGelKLPxczawtcDuS6+2GEWRFq04wHjwIDSu3LA153907A64ntKqekkEHuvtrd5yQeryf8p28b\nb1TxMbN2wGnAX+KOJW5m1gI4ljD1C+6+xd2/iTeq2NUDGicmy2wCrIo5noxx95mEqX9Spa4/8xjw\n8yiuraQQk8TSo72Bd+ONJFZ3A78FKrD8eI3VEVgDPJKoTvuLmTWNO6i4uPtK4A5gBbAaWOfuf4s3\nqtjt6+6rE48/A/aN4iJKCjEws2bAs8CV7v5t3PHEwcwGAV+4+/txx5Il6gF9gPvcvTewkYiqB6qD\nRH35EEKybAM0NbNz4o0qeyRmk45kPIGSQoaZWX1CQsh39/+LO54Y9QMGm1khYanWE83syXhDilUR\nUOTuJSXHZwhJorY6GfjE3de4+1bg/4CjY44pbp+b2f4AifsvoriIkkIGJdaffghY5O4T4o4nTu7+\nO3dv5+45hAbEN9y91v4SdPfPgE/N7NDErpOAhTGGFLcVwJFm1iTx/+YkanHDe0Lq+jPnAc9HcREl\nhczqB5xL+FU8N3E7Ne6gJGtcBuSb2XygF3BrzPHEJlFiegaYA3xI+K6qNVNemNlTwNvAoWZWZGYX\nArcBp5jZEkJJ6rZIrq1pLkREpIRKCiIikqSkICIiSUoKIiKSpKQgIiJJSgoiIpKkpCCSYGY/pHQV\nnmtmVTai2MxyUme8FMlW9eIOQCSLbHb3XnEHIRInlRREdsHMCs3sdjP70MzeM7ODE/tzzOwNM5tv\nZq+bWfvE/n3NbJqZzUvcSqZnqGtmDybWCPibmTVOHH95Yo2N+WY2Naa3KQIoKYikalyq+ujslOfW\nuXt34L8Js7sC3As85u49gHxgYmL/ROD/c/eehPmLFiT2dwImuXs34BvgrMT+PKB34jy/ierNiaRD\nI5pFEsxsg7s3K2N/IXCiuy9LTGj4mbu3NLMvgf3dfWti/2p3b2Vma4B27v59yjlygNcSC6RgZtcD\n9d19rJm9AmwAngOec/cNEb9VkZ1SSUEkPb6TxxXxfcrjH9jepncaMIlQqpidWFRGJBZKCiLpOTvl\n/u3E41lsXyJyJPDPxOPXgdGQXIO6xc5OamZ1gAPcfQZwPdAC+FFpRSRT9ItEZLvGZjY3ZfsVdy/p\nlrpXYvbS74ERiX2XEVZKu46watoFif1XAJMTM1v+QEgQqylbXeDJROIwYKKW4ZQ4qU1BZBcSbQq5\n7v5l3LGIRE3VRyIikqSSgoiIJKmkICIiSUoKIiKSpKQgIiJJSgoiIpKkpCAiIkn/P1YEHoYdRnls\nAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U_MNcO9JttAR",
        "colab_type": "text"
      },
      "source": [
        "The super-naive Chapter 3 baseline got us to a test accuracy of ~88%, so the small RNN doesn't do well compared to this baseline. Our inputs only consider the first 500 words, so our model has less information than the baseline. In addition, `SimpleRNN` isn't good at long sequences. Let's try something fancier.\n",
        "\n",
        "### LSTM and GRU\n",
        "\n",
        "The LSTM is an attempt to solve the vanishing gradient problem, which makes long-term dependencies in our data difficult to learn. It adds a \"carry layer\" c_t to our data flow (in addition to state, and input)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HgT0-NT7uW7H",
        "colab_type": "code",
        "outputId": "44aa610f-81fc-4518-8424-6ad79000b6dd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 680
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_features, 32))\n",
        "model.add(keras.layers.LSTM(32))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "\n",
        "history = model.fit(input_train, y_train, epochs=10, batch_size=128, validation_split=0.2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0619 09:02:09.064861 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "W0619 09:02:09.112024 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "W0619 09:02:09.119617 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "W0619 09:02:09.381532 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n",
            "W0619 09:02:09.404931 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3376: The name tf.log is deprecated. Please use tf.math.log instead.\n",
            "\n",
            "W0619 09:02:09.414445 140188709402496 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/nn_impl.py:180: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
            "W0619 09:02:10.427320 140188709402496 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:986: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Train on 20000 samples, validate on 5000 samples\n",
            "Epoch 1/10\n",
            "20000/20000 [==============================] - 144s 7ms/step - loss: 0.5016 - acc: 0.7665 - val_loss: 0.3547 - val_acc: 0.8580\n",
            "Epoch 2/10\n",
            "20000/20000 [==============================] - 139s 7ms/step - loss: 0.2936 - acc: 0.8839 - val_loss: 0.3124 - val_acc: 0.8656\n",
            "Epoch 3/10\n",
            "20000/20000 [==============================] - 140s 7ms/step - loss: 0.2349 - acc: 0.9101 - val_loss: 0.3530 - val_acc: 0.8606\n",
            "Epoch 4/10\n",
            "20000/20000 [==============================] - 139s 7ms/step - loss: 0.1983 - acc: 0.9267 - val_loss: 0.2999 - val_acc: 0.8756\n",
            "Epoch 5/10\n",
            "20000/20000 [==============================] - 140s 7ms/step - loss: 0.1775 - acc: 0.9344 - val_loss: 0.2896 - val_acc: 0.8788\n",
            "Epoch 6/10\n",
            "20000/20000 [==============================] - 139s 7ms/step - loss: 0.1608 - acc: 0.9422 - val_loss: 0.3417 - val_acc: 0.8646\n",
            "Epoch 7/10\n",
            "20000/20000 [==============================] - 139s 7ms/step - loss: 0.1398 - acc: 0.9502 - val_loss: 0.3249 - val_acc: 0.8846\n",
            "Epoch 8/10\n",
            "20000/20000 [==============================] - 140s 7ms/step - loss: 0.1319 - acc: 0.9546 - val_loss: 0.4744 - val_acc: 0.8650\n",
            "Epoch 9/10\n",
            "20000/20000 [==============================] - 139s 7ms/step - loss: 0.1259 - acc: 0.9552 - val_loss: 0.3529 - val_acc: 0.8584\n",
            "Epoch 10/10\n",
            "20000/20000 [==============================] - 140s 7ms/step - loss: 0.1113 - acc: 0.9618 - val_loss: 0.3938 - val_acc: 0.8814\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_g7apqTx9_j6",
        "colab_type": "code",
        "outputId": "f3ffb48b-f36e-49b2-a6e2-e39f0b44f731",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "plotLossAndAccuracy(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nE5ElC2OMMRFZsjDGGBPR/wMYDzDz6uKoHAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HOohxvpb-Woa",
        "colab_type": "text"
      },
      "source": [
        "The accuracy jumps up significantly, compared to the `SimpleRNN` network. It doesn't do better because analysing the global, long-term structure of the reviews (which the LSTM is good at) isn't a great way of doing sentiment analysis - word frequency analysis does better. But for more complicated problems, it's helpful.\n",
        "\n",
        "## Advanced use of recurrent neural networks\n",
        "\n",
        "Three advanced techniques for improving the performance/generalisability of RNNs, demonstrated on a temperature forecasting problem."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K9MhAWRV-fUo",
        "colab_type": "code",
        "outputId": "395b620b-1fd6-4650-94b1-f87680aca853",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        }
      },
      "source": [
        "# First, download the raw  data  (14 different atmospheric quantities + temperature over 7 years).\n",
        "!wget --no-check-certificate \\\n",
        "    https://s3.amazonaws.com/keras-datasets/jena_climate_2009_2016.csv.zip \\\n",
        "    -O /tmp/climatedata.zip\n",
        "zip_ref = zipfile.ZipFile(\"/tmp/climatedata.zip\", 'r')\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-06-20 12:27:50--  https://s3.amazonaws.com/keras-datasets/jena_climate_2009_2016.csv.zip\n",
            "Resolving s3.amazonaws.com (s3.amazonaws.com)... 52.216.85.165\n",
            "Connecting to s3.amazonaws.com (s3.amazonaws.com)|52.216.85.165|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 13568290 (13M) [application/zip]\n",
            "Saving to: ‘/tmp/climatedata.zip’\n",
            "\n",
            "/tmp/climatedata.zi 100%[===================>]  12.94M  15.8MB/s    in 0.8s    \n",
            "\n",
            "2019-06-20 12:27:51 (15.8 MB/s) - ‘/tmp/climatedata.zip’ saved [13568290/13568290]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5X3WwEke_6X0",
        "colab_type": "code",
        "outputId": "a19bc472-7414-434f-a5ea-a0df1ca74f53",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        }
      },
      "source": [
        "fname = \"/tmp/jena_climate_2009_2016.csv\"\n",
        "\n",
        "# Chollet processes this into a Numpy array manually, but here we'll use Pandas\n",
        "\n",
        "df = pd.read_csv(fname, index_col=0, parse_dates=True)\n",
        "\n",
        "df.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>p (mbar)</th>\n",
              "      <th>T (degC)</th>\n",
              "      <th>Tpot (K)</th>\n",
              "      <th>Tdew (degC)</th>\n",
              "      <th>rh (%)</th>\n",
              "      <th>VPmax (mbar)</th>\n",
              "      <th>VPact (mbar)</th>\n",
              "      <th>VPdef (mbar)</th>\n",
              "      <th>sh (g/kg)</th>\n",
              "      <th>H2OC (mmol/mol)</th>\n",
              "      <th>rho (g/m**3)</th>\n",
              "      <th>wv (m/s)</th>\n",
              "      <th>max. wv (m/s)</th>\n",
              "      <th>wd (deg)</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Date Time</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>2009-01-01 00:10:00</th>\n",
              "      <td>996.52</td>\n",
              "      <td>-8.02</td>\n",
              "      <td>265.40</td>\n",
              "      <td>-8.90</td>\n",
              "      <td>93.3</td>\n",
              "      <td>3.33</td>\n",
              "      <td>3.11</td>\n",
              "      <td>0.22</td>\n",
              "      <td>1.94</td>\n",
              "      <td>3.12</td>\n",
              "      <td>1307.75</td>\n",
              "      <td>1.03</td>\n",
              "      <td>1.75</td>\n",
              "      <td>152.3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2009-01-01 00:20:00</th>\n",
              "      <td>996.57</td>\n",
              "      <td>-8.41</td>\n",
              "      <td>265.01</td>\n",
              "      <td>-9.28</td>\n",
              "      <td>93.4</td>\n",
              "      <td>3.23</td>\n",
              "      <td>3.02</td>\n",
              "      <td>0.21</td>\n",
              "      <td>1.89</td>\n",
              "      <td>3.03</td>\n",
              "      <td>1309.80</td>\n",
              "      <td>0.72</td>\n",
              "      <td>1.50</td>\n",
              "      <td>136.1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2009-01-01 00:30:00</th>\n",
              "      <td>996.53</td>\n",
              "      <td>-8.51</td>\n",
              "      <td>264.91</td>\n",
              "      <td>-9.31</td>\n",
              "      <td>93.9</td>\n",
              "      <td>3.21</td>\n",
              "      <td>3.01</td>\n",
              "      <td>0.20</td>\n",
              "      <td>1.88</td>\n",
              "      <td>3.02</td>\n",
              "      <td>1310.24</td>\n",
              "      <td>0.19</td>\n",
              "      <td>0.63</td>\n",
              "      <td>171.6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2009-01-01 00:40:00</th>\n",
              "      <td>996.51</td>\n",
              "      <td>-8.31</td>\n",
              "      <td>265.12</td>\n",
              "      <td>-9.07</td>\n",
              "      <td>94.2</td>\n",
              "      <td>3.26</td>\n",
              "      <td>3.07</td>\n",
              "      <td>0.19</td>\n",
              "      <td>1.92</td>\n",
              "      <td>3.08</td>\n",
              "      <td>1309.19</td>\n",
              "      <td>0.34</td>\n",
              "      <td>0.50</td>\n",
              "      <td>198.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2009-01-01 00:50:00</th>\n",
              "      <td>996.51</td>\n",
              "      <td>-8.27</td>\n",
              "      <td>265.15</td>\n",
              "      <td>-9.04</td>\n",
              "      <td>94.1</td>\n",
              "      <td>3.27</td>\n",
              "      <td>3.08</td>\n",
              "      <td>0.19</td>\n",
              "      <td>1.92</td>\n",
              "      <td>3.09</td>\n",
              "      <td>1309.00</td>\n",
              "      <td>0.32</td>\n",
              "      <td>0.63</td>\n",
              "      <td>214.3</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                     p (mbar)  T (degC)  ...  max. wv (m/s)  wd (deg)\n",
              "Date Time                                ...                         \n",
              "2009-01-01 00:10:00    996.52     -8.02  ...           1.75     152.3\n",
              "2009-01-01 00:20:00    996.57     -8.41  ...           1.50     136.1\n",
              "2009-01-01 00:30:00    996.53     -8.51  ...           0.63     171.6\n",
              "2009-01-01 00:40:00    996.51     -8.31  ...           0.50     198.0\n",
              "2009-01-01 00:50:00    996.51     -8.27  ...           0.63     214.3\n",
              "\n",
              "[5 rows x 14 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eT9zzVlkAWUL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Convert to a Numpy array\n",
        "float_data = df.to_numpy()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "stX5FybLFvzz",
        "colab_type": "code",
        "outputId": "90e835aa-9c54-4dcc-8f52-0343a74ec8f2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 291
        }
      },
      "source": [
        "df[\"T (degC)\"].plot()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f7f98748eb8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 43
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEACAYAAACwB81wAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Hzja-0lbF5L5",
        "colab_type": "code",
        "outputId": "d2dfdb44-8162-4da4-d987-43f709c7e02f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 301
        }
      },
      "source": [
        "# Because it's a time series, we can do nice things like downsampling\n",
        "df[\"T (degC)\"].resample('M').mean().plot()\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f7f959869e8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 44
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAEKCAYAAAD6q1UVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAMsaJEm8lZ3BAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2UTVRSc8HAlD",
        "colab_type": "code",
        "outputId": "23a4bbb1-c30a-41b2-d425-38cfad131ffd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 324
        }
      },
      "source": [
        "# There's clear seasonality, but can we predict on a daily scale?\n",
        "df[\"T (degC)\"].resample(\"H\").mean().iloc[:100].plot()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f7fec6787b8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 52
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEhCAYAAAB89xvjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngS8iEiYU+CIiYUKBLyISJv4/BZfOc7m4TkoAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GRHlY-j1HM-0",
        "colab_type": "text"
      },
      "source": [
        "### Preparing the data\n",
        "\n",
        "Goal - given data going as far back as `lookback` timesteps and sampled every `steps` timesteps, can you predict the temperature in `delay` timesteps?\n",
        "Use:\n",
        "- `lookback=720` (i.e. 5 days)\n",
        "- `steps=6` (one data point per hour)\n",
        "- `delay=144` (target is 24 hours in the future)\n",
        "\n",
        "Steps:\n",
        "-  Preprocess the data so a neural network can ingest it: it's already vectorized, just needs normalization.\n",
        "- Write a Python generator that takes the current array of float data and yields batchs from the recent past, along with a target temperature in the future.\n",
        "\n",
        "We'll use the first 200,000 timesteps as training data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S_rSd_uOeE5h",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Normalize\n",
        "mean = float_data[:200000].mean(axis=0)\n",
        "float_data -= mean\n",
        "std = float_data[:200000].std(axis=0)\n",
        "float_data /= std"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "948sYQkPeGbp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Generator yielding timeseries samples and their targets\n",
        "\n",
        "def generator(data, lookback, delay, min_index, max_index, shuffle=False, batch_size=128, step=6):\n",
        "  \"\"\"\n",
        "  Parameters\n",
        "  ----------\n",
        "  data : 2d numpy array\n",
        "      The array of floating-point timeseries data, normalized.\n",
        "  lookback : int\n",
        "      How many timesteps back the input data should go.\n",
        "  delay: int\n",
        "      How many timesteps in the future the target should be.\n",
        "  min_index: int\n",
        "      The lower bound of timesteps to draw from (helps keep validation/test data separate) \n",
        "  max_index: int\n",
        "      The upper bound of timesteps to draw from (helps keep validation/test data separate)\n",
        "  shuffle: bool\n",
        "      Whether to shuffle the samples or draw them in chronological order\n",
        "  batch_size: int\n",
        "      The number of samples per batch.\n",
        "  step: int\n",
        "      The period, in timesteps, at which you sample data.\n",
        "  \"\"\"\n",
        "  if max_index is None:\n",
        "    max_index = len(data) - delay - 1\n",
        "  i = min_index + lookback\n",
        "  while True:\n",
        "    if shuffle:\n",
        "      rows = np.random.randint(min_index+lookback, max_index, size=batch_size)\n",
        "    else:\n",
        "      if i+batch_size >= max_index:\n",
        "        i = min_index + lookback\n",
        "      rows = np.arange(i, min(i+batch_size, max_index))\n",
        "      i += len(rows)\n",
        "      \n",
        "    samples = np.zeros((len(rows), lookback // step, data.shape[-1]))\n",
        "    targets = np.zeros((len(rows),))\n",
        "    for j, row in enumerate(rows):\n",
        "      indices = range(rows[j] - lookback, rows[j], step)\n",
        "      samples[j] = data[indices]\n",
        "      targets[j] = data[rows[j] + delay][1]  # temperature is the second column\n",
        "    yield samples, targets\n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ATbGvOi_hB37",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Create a training, validation, and test generator looking at different segments of the data\n",
        "lookback = 1440\n",
        "step = 6\n",
        "delay = 144\n",
        "batch_size = 128\n",
        "\n",
        "train_gen = generator(float_data,\n",
        "                      lookback=lookback,\n",
        "                      delay=delay,\n",
        "                      min_index=0,\n",
        "                      max_index=200000,\n",
        "                      shuffle=True,\n",
        "                      step=step,\n",
        "                      batch_size=batch_size\n",
        "                     )\n",
        "val_gen = generator(float_data,\n",
        "                    lookback=lookback,\n",
        "                    delay=delay,\n",
        "                    min_index=200001,\n",
        "                    max_index=300000,\n",
        "                    step=step,\n",
        "                    batch_size=batch_size\n",
        "                   )\n",
        "test_gen = generator(float_data,\n",
        "                     lookback=lookback,\n",
        "                     delay=delay,\n",
        "                     min_index=300001,\n",
        "                     max_index=None,\n",
        "                     step=step,\n",
        "                     batch_size=batch_size\n",
        "                    )\n",
        "\n",
        "# How many steps to draw from val_gen in order to see the entire validation set\n",
        "val_steps = (300000 - 200001 - lookback) // batch_size\n",
        "# As above, for the test set\n",
        "test_steps = (len(float_data) - 300001 - lookback) // batch_size\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4BG8aLE7i151",
        "colab_type": "text"
      },
      "source": [
        "### A non-ML baseline\n",
        "\n",
        "As a sanity check, we will predict that the temperature in 24 hours is always equal to the current temperature, and evaluate this with MAE"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nnpafVY3jLCF",
        "colab_type": "code",
        "outputId": "ed7caea8-cc93-44cb-f689-597d0a70acf5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "def evaluate_naive_method():\n",
        "  batch_maes = []\n",
        "  for step in range(val_steps):\n",
        "    samples, targets = next(val_gen)\n",
        "    preds = samples[:, -1, 1]\n",
        "    mae = np.mean(np.abs(preds-targets))\n",
        "    batch_maes.append(mae)\n",
        "  print(f\"{np.mean(batch_maes):.3f}\")\n",
        "  return np.mean(batch_maes)\n",
        "  \n",
        "  \n",
        "naive_mae = evaluate_naive_method()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.290\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7S8DAQCNjjKw",
        "colab_type": "code",
        "outputId": "de3a6872-8698-4720-9aea-4c791f59180a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "# The MAE of 0.29 isn't really interpretable, as it's on normalised data.\n",
        "# Multiply by the standard deviation of temperature\n",
        "temperature_std = pd.read_csv(fname, index_col=0)[\"T (degC)\"].iloc[:200000].std()\n",
        "\n",
        "print(f\"Naive MAE in degrees: {temperature_std*naive_mae:.2f}\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Naive MAE in degrees: 2.57\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BQ3ByxQdkyEp",
        "colab_type": "text"
      },
      "source": [
        "### Basic ML approach\n",
        "\n",
        "Just flatten the data and run it through two `Dense` layers"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-WKIvWA3k43a",
        "colab_type": "code",
        "outputId": "f9a51312-9712-4e42-ea9f-340ad6fe8f3d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1010
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Flatten(input_shape=(lookback // step, float_data.shape[-1])))\n",
        "model.add(keras.layers.Dense(32, activation=\"relu\"))\n",
        "model.add(keras.layers.Dense(1))  # Note no activation on this layer\n",
        "\n",
        "model.compile(optimizer=keras.optimizers.RMSprop(), loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(train_gen,\n",
        "                             steps_per_epoch=500,\n",
        "                             epochs=20,\n",
        "                             validation_data=val_gen,\n",
        "                             validation_steps=val_steps\n",
        "                             )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0620 12:29:35.197392 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "W0620 12:29:35.244894 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "W0620 12:29:35.269035 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "W0620 12:29:35.319800 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n",
            "W0620 12:29:35.493102 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:986: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
            "\n",
            "W0620 12:29:35.507735 140290881169280 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:973: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "500/500 [==============================] - 19s 37ms/step - loss: 2.1142 - val_loss: 1.1346\n",
            "Epoch 2/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.8107 - val_loss: 0.4233\n",
            "Epoch 3/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.3761 - val_loss: 0.3189\n",
            "Epoch 4/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2820 - val_loss: 0.3132\n",
            "Epoch 5/20\n",
            "500/500 [==============================] - 16s 31ms/step - loss: 0.2626 - val_loss: 0.3523\n",
            "Epoch 6/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2477 - val_loss: 0.3386\n",
            "Epoch 7/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2423 - val_loss: 0.3176\n",
            "Epoch 8/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2335 - val_loss: 0.3188\n",
            "Epoch 9/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2293 - val_loss: 0.3300\n",
            "Epoch 10/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2240 - val_loss: 0.3238\n",
            "Epoch 11/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2207 - val_loss: 0.3372\n",
            "Epoch 12/20\n",
            "500/500 [==============================] - 16s 32ms/step - loss: 0.2166 - val_loss: 0.3215\n",
            "Epoch 13/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2140 - val_loss: 0.3323\n",
            "Epoch 14/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2112 - val_loss: 0.3565\n",
            "Epoch 15/20\n",
            "500/500 [==============================] - 16s 31ms/step - loss: 0.2097 - val_loss: 0.3442\n",
            "Epoch 16/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2074 - val_loss: 0.3444\n",
            "Epoch 17/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2048 - val_loss: 0.3331\n",
            "Epoch 18/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2027 - val_loss: 0.3373\n",
            "Epoch 19/20\n",
            "500/500 [==============================] - 15s 30ms/step - loss: 0.2006 - val_loss: 0.3550\n",
            "Epoch 20/20\n",
            "500/500 [==============================] - 15s 31ms/step - loss: 0.2007 - val_loss: 0.3478\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RhAUh9ITmbxU",
        "colab_type": "code",
        "outputId": "ab61d118-205c-4b83-bc41-a8bd70e787fb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "def plotLoss(history):\n",
        "  history_dict = history.history\n",
        "  loss = history_dict[\"loss\"]\n",
        "  val_loss = history_dict[\"val_loss\"]\n",
        "  epochs = range(1, len(loss) +1)\n",
        "\n",
        "  plt.plot(epochs, loss, \"bo\", label=\"Training Loss\")\n",
        "  plt.plot(epochs, val_loss, \"b\", label=\"Validation Loss\")\n",
        "  plt.title(\"Training and validation loss\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Loss\")\n",
        "  plt.legend()\n",
        "#   plt.xlim(2,)\n",
        "#   plt.ylim(0,0.5)\n",
        "  plt.show()\n",
        "\n",
        "plotLoss(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nVwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Q4wuWBgpsTPt",
        "colab_type": "text"
      },
      "source": [
        "Actually, we can't do better than the common-sense baseline with this method!\n",
        "\n",
        "### First recurrent baseline\n",
        "\n",
        "The first model flattened everything, removing any notion of time. Let's try a recurrent model instead, using the Gated Recurrent Unit (GRU) layer. (Like LSTM, cheaper but perhaps with less representational power)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DLsN9pr7vDFO",
        "colab_type": "code",
        "outputId": "7d3f3c2f-8061-434a-ab96-bf9b14036844",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 752
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.GRU(32, input_shape=(None, float_data.shape[-1])))\n",
        "model.add(keras.layers.Dense(1))\n",
        "\n",
        "model.compile(optimizer=keras.optimizers.RMSprop(), loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(train_gen,\n",
        "                             steps_per_epoch=500,\n",
        "                             epochs=20,\n",
        "                             validation_data=val_gen,\n",
        "                             validation_steps=val_steps\n",
        "                             )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "500/500 [==============================] - 220s 439ms/step - loss: 0.3118 - val_loss: 0.2623\n",
            "Epoch 2/20\n",
            "500/500 [==============================] - 218s 436ms/step - loss: 0.2835 - val_loss: 0.2661\n",
            "Epoch 3/20\n",
            "500/500 [==============================] - 218s 437ms/step - loss: 0.2771 - val_loss: 0.2679\n",
            "Epoch 4/20\n",
            "500/500 [==============================] - 218s 437ms/step - loss: 0.2721 - val_loss: 0.2749\n",
            "Epoch 5/20\n",
            "500/500 [==============================] - 220s 439ms/step - loss: 0.2668 - val_loss: 0.2625\n",
            "Epoch 6/20\n",
            "500/500 [==============================] - 220s 439ms/step - loss: 0.2616 - val_loss: 0.2781\n",
            "Epoch 7/20\n",
            "500/500 [==============================] - 220s 441ms/step - loss: 0.2563 - val_loss: 0.2743\n",
            "Epoch 8/20\n",
            "500/500 [==============================] - 220s 440ms/step - loss: 0.2499 - val_loss: 0.2716\n",
            "Epoch 9/20\n",
            "500/500 [==============================] - 220s 439ms/step - loss: 0.2438 - val_loss: 0.2760\n",
            "Epoch 10/20\n",
            "500/500 [==============================] - 221s 442ms/step - loss: 0.2406 - val_loss: 0.2703\n",
            "Epoch 11/20\n",
            "500/500 [==============================] - 218s 436ms/step - loss: 0.2352 - val_loss: 0.2787\n",
            "Epoch 12/20\n",
            "500/500 [==============================] - 217s 433ms/step - loss: 0.2327 - val_loss: 0.2778\n",
            "Epoch 13/20\n",
            "500/500 [==============================] - 216s 433ms/step - loss: 0.2301 - val_loss: 0.2781\n",
            "Epoch 14/20\n",
            "500/500 [==============================] - 217s 434ms/step - loss: 0.2237 - val_loss: 0.2875\n",
            "Epoch 15/20\n",
            "500/500 [==============================] - 215s 431ms/step - loss: 0.2240 - val_loss: 0.2842\n",
            "Epoch 16/20\n",
            "500/500 [==============================] - 216s 432ms/step - loss: 0.2181 - val_loss: 0.2927\n",
            "Epoch 17/20\n",
            "500/500 [==============================] - 217s 433ms/step - loss: 0.2155 - val_loss: 0.2932\n",
            "Epoch 18/20\n",
            "500/500 [==============================] - 214s 429ms/step - loss: 0.2134 - val_loss: 0.2973\n",
            "Epoch 19/20\n",
            "500/500 [==============================] - 213s 426ms/step - loss: 0.2095 - val_loss: 0.2979\n",
            "Epoch 20/20\n",
            "500/500 [==============================] - 213s 427ms/step - loss: 0.2072 - val_loss: 0.3019\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SrF3xiRkvn69",
        "colab_type": "code",
        "outputId": "4d1d3694-f4f7-474c-8110-2f36ceddd9e4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "plotLoss(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Ou71unNvsQz",
        "colab_type": "text"
      },
      "source": [
        "### Using recurrent dropout to prevent overfitting\n",
        "\n",
        "The loss curves show clear overfitting. But naive Dropout won't help - instead, we need the same dropout mask applied at every timestep (and should be applied to the inner recurrent activations of the GRU and LSTM layers). "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DLv2izz5wXln",
        "colab_type": "code",
        "outputId": "cd3ed166-5668-4ed3-e60f-e0351f3adc36",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1478
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.GRU(32,\n",
        "                           dropout=0.2,\n",
        "                           recurrent_dropout=0.2,\n",
        "                           input_shape=(None, float_data.shape[-1])))\n",
        "model.add(keras.layers.Dense(1))\n",
        "\n",
        "model.compile(optimizer=keras.optimizers.RMSprop(), loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(train_gen,\n",
        "                             steps_per_epoch=500,\n",
        "                             epochs=40,  # Twice as many epochs, as dropout makes convergence slower\n",
        "                             validation_data=val_gen,\n",
        "                             validation_steps=val_steps\n",
        "                             )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0619 13:50:51.928678 140188709402496 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/40\n",
            "500/500 [==============================] - 249s 497ms/step - loss: 0.3526 - val_loss: 0.2767\n",
            "Epoch 2/40\n",
            "500/500 [==============================] - 247s 494ms/step - loss: 0.3151 - val_loss: 0.2720\n",
            "Epoch 3/40\n",
            "500/500 [==============================] - 247s 495ms/step - loss: 0.3088 - val_loss: 0.2718\n",
            "Epoch 4/40\n",
            "500/500 [==============================] - 247s 495ms/step - loss: 0.3040 - val_loss: 0.2684\n",
            "Epoch 5/40\n",
            "500/500 [==============================] - 247s 495ms/step - loss: 0.3016 - val_loss: 0.2650\n",
            "Epoch 6/40\n",
            "500/500 [==============================] - 247s 495ms/step - loss: 0.2981 - val_loss: 0.2692\n",
            "Epoch 7/40\n",
            "500/500 [==============================] - 247s 495ms/step - loss: 0.2961 - val_loss: 0.2679\n",
            "Epoch 8/40\n",
            "500/500 [==============================] - 247s 493ms/step - loss: 0.2926 - val_loss: 0.2716\n",
            "Epoch 9/40\n",
            "500/500 [==============================] - 247s 494ms/step - loss: 0.2933 - val_loss: 0.2631\n",
            "Epoch 10/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2912 - val_loss: 0.2642\n",
            "Epoch 11/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2892 - val_loss: 0.2645\n",
            "Epoch 12/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2881 - val_loss: 0.2658\n",
            "Epoch 13/40\n",
            "500/500 [==============================] - 246s 491ms/step - loss: 0.2881 - val_loss: 0.2679\n",
            "Epoch 14/40\n",
            "500/500 [==============================] - 246s 493ms/step - loss: 0.2860 - val_loss: 0.2659\n",
            "Epoch 15/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2855 - val_loss: 0.2641\n",
            "Epoch 16/40\n",
            "500/500 [==============================] - 247s 493ms/step - loss: 0.2834 - val_loss: 0.2665\n",
            "Epoch 17/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2843 - val_loss: 0.2674\n",
            "Epoch 18/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2820 - val_loss: 0.2645\n",
            "Epoch 19/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2825 - val_loss: 0.2634\n",
            "Epoch 20/40\n",
            "500/500 [==============================] - 246s 493ms/step - loss: 0.2793 - val_loss: 0.2680\n",
            "Epoch 21/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2794 - val_loss: 0.2716\n",
            "Epoch 22/40\n",
            "500/500 [==============================] - 246s 491ms/step - loss: 0.2761 - val_loss: 0.2678\n",
            "Epoch 23/40\n",
            "500/500 [==============================] - 245s 490ms/step - loss: 0.2789 - val_loss: 0.2622\n",
            "Epoch 24/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2762 - val_loss: 0.2611\n",
            "Epoch 25/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2763 - val_loss: 0.2652\n",
            "Epoch 26/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2750 - val_loss: 0.2650\n",
            "Epoch 27/40\n",
            "500/500 [==============================] - 246s 491ms/step - loss: 0.2765 - val_loss: 0.2686\n",
            "Epoch 28/40\n",
            "500/500 [==============================] - 246s 491ms/step - loss: 0.2749 - val_loss: 0.2617\n",
            "Epoch 29/40\n",
            "500/500 [==============================] - 247s 494ms/step - loss: 0.2754 - val_loss: 0.2667\n",
            "Epoch 30/40\n",
            "500/500 [==============================] - 246s 493ms/step - loss: 0.2723 - val_loss: 0.2642\n",
            "Epoch 31/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2752 - val_loss: 0.2673\n",
            "Epoch 32/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2713 - val_loss: 0.2678\n",
            "Epoch 33/40\n",
            "500/500 [==============================] - 246s 491ms/step - loss: 0.2704 - val_loss: 0.2628\n",
            "Epoch 34/40\n",
            "500/500 [==============================] - 246s 493ms/step - loss: 0.2702 - val_loss: 0.2635\n",
            "Epoch 35/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2712 - val_loss: 0.2650\n",
            "Epoch 36/40\n",
            "500/500 [==============================] - 246s 492ms/step - loss: 0.2697 - val_loss: 0.2623\n",
            "Epoch 37/40\n",
            "500/500 [==============================] - 250s 501ms/step - loss: 0.2694 - val_loss: 0.2681\n",
            "Epoch 38/40\n",
            "500/500 [==============================] - 249s 497ms/step - loss: 0.2689 - val_loss: 0.2703\n",
            "Epoch 39/40\n",
            "500/500 [==============================] - 248s 496ms/step - loss: 0.2688 - val_loss: 0.2611\n",
            "Epoch 40/40\n",
            "500/500 [==============================] - 248s 496ms/step - loss: 0.2683 - val_loss: 0.2753\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iIhI5UXawpUQ",
        "colab_type": "code",
        "outputId": "8a53374c-5c6c-41b7-dd75-95c65e3763c7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "plotLoss(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TaIWP8Eiwq2w",
        "colab_type": "text"
      },
      "source": [
        "Recurrent dropout successfully prevents overfitting, and the validation scores have now stabilised. But the scores aren't improving much!\n",
        "\n",
        "### Stacking recurrent layers\n",
        "\n",
        "We've not overfitting, but we've hit a performance bottleneck - time to increase the model capacity. Let's stack some GRUs."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XPXdCW_IBIkN",
        "colab_type": "code",
        "outputId": "bd49754b-9983-4af8-de77-f6f459e29126",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1531
        }
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.GRU(32,\n",
        "                           dropout=0.1,\n",
        "                           recurrent_dropout=0.5,\n",
        "                           return_sequences=True,\n",
        "                           input_shape=(None, float_data.shape[-1])\n",
        "                          ))\n",
        "model.add(keras.layers.GRU(64,\n",
        "                           activation=\"relu\",\n",
        "                           dropout=0.1,\n",
        "                           recurrent_dropout=0.5,\n",
        "                           ))\n",
        "\n",
        "model.add(keras.layers.Dense(1))\n",
        "\n",
        "model.compile(optimizer=keras.optimizers.RMSprop(), loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(train_gen,\n",
        "                             steps_per_epoch=500,\n",
        "                             epochs=40,  # Twice as many epochs, as dropout makes convergence slower\n",
        "                             validation_data=val_gen,\n",
        "                             validation_steps=val_steps\n",
        "                             )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0619 21:31:29.914033 139653638489984 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "W0619 21:31:30.739833 139653638489984 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/40\n",
            "500/500 [==============================] - 405s 809ms/step - loss: 0.3349 - val_loss: 0.2785\n",
            "Epoch 2/40\n",
            "500/500 [==============================] - 403s 805ms/step - loss: 0.3123 - val_loss: 0.2722\n",
            "Epoch 3/40\n",
            "500/500 [==============================] - 401s 802ms/step - loss: 0.3069 - val_loss: 0.2724\n",
            "Epoch 4/40\n",
            "500/500 [==============================] - 400s 800ms/step - loss: 0.3018 - val_loss: 0.2699\n",
            "Epoch 5/40\n",
            "500/500 [==============================] - 398s 796ms/step - loss: 0.2969 - val_loss: 0.2678\n",
            "Epoch 6/40\n",
            "500/500 [==============================] - 400s 799ms/step - loss: 0.2953 - val_loss: 0.2762\n",
            "Epoch 7/40\n",
            "500/500 [==============================] - 398s 796ms/step - loss: 0.2931 - val_loss: 0.2696\n",
            "Epoch 8/40\n",
            "500/500 [==============================] - 396s 792ms/step - loss: 0.2892 - val_loss: 0.2733\n",
            "Epoch 9/40\n",
            "500/500 [==============================] - 396s 792ms/step - loss: 0.2881 - val_loss: 0.2691\n",
            "Epoch 10/40\n",
            "500/500 [==============================] - 396s 792ms/step - loss: 0.2846 - val_loss: 0.2702\n",
            "Epoch 11/40\n",
            "500/500 [==============================] - 396s 792ms/step - loss: 0.2823 - val_loss: 0.2708\n",
            "Epoch 12/40\n",
            "500/500 [==============================] - 396s 793ms/step - loss: 0.2795 - val_loss: 0.2675\n",
            "Epoch 13/40\n",
            "500/500 [==============================] - 396s 791ms/step - loss: 0.2785 - val_loss: 0.2694\n",
            "Epoch 14/40\n",
            "500/500 [==============================] - 394s 789ms/step - loss: 0.2763 - val_loss: 0.2599\n",
            "Epoch 15/40\n",
            "500/500 [==============================] - 395s 790ms/step - loss: 0.2725 - val_loss: 0.2695\n",
            "Epoch 16/40\n",
            "500/500 [==============================] - 394s 787ms/step - loss: 0.2719 - val_loss: 0.2672\n",
            "Epoch 17/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2737 - val_loss: 0.2662\n",
            "Epoch 18/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2696 - val_loss: 0.2667\n",
            "Epoch 19/40\n",
            "500/500 [==============================] - 393s 786ms/step - loss: 0.2686 - val_loss: 0.2656\n",
            "Epoch 20/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2687 - val_loss: 0.2660\n",
            "Epoch 21/40\n",
            "500/500 [==============================] - 394s 787ms/step - loss: 0.2656 - val_loss: 0.2798\n",
            "Epoch 22/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2658 - val_loss: 0.2688\n",
            "Epoch 23/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2641 - val_loss: 0.2661\n",
            "Epoch 24/40\n",
            "500/500 [==============================] - 394s 789ms/step - loss: 0.2636 - val_loss: 0.2647\n",
            "Epoch 25/40\n",
            "500/500 [==============================] - 395s 790ms/step - loss: 0.2625 - val_loss: 0.2696\n",
            "Epoch 26/40\n",
            "500/500 [==============================] - 393s 787ms/step - loss: 0.2607 - val_loss: 0.2680\n",
            "Epoch 27/40\n",
            "500/500 [==============================] - 396s 792ms/step - loss: 0.2623 - val_loss: 0.2744\n",
            "Epoch 28/40\n",
            "500/500 [==============================] - 403s 807ms/step - loss: 0.2624 - val_loss: 0.2885\n",
            "Epoch 29/40\n",
            "500/500 [==============================] - 393s 787ms/step - loss: 0.2602 - val_loss: 0.2687\n",
            "Epoch 30/40\n",
            "500/500 [==============================] - 392s 783ms/step - loss: 0.2581 - val_loss: 0.2683\n",
            "Epoch 31/40\n",
            "500/500 [==============================] - 393s 787ms/step - loss: 0.2593 - val_loss: 0.2716\n",
            "Epoch 32/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2587 - val_loss: 0.2688\n",
            "Epoch 33/40\n",
            "500/500 [==============================] - 393s 787ms/step - loss: 0.2579 - val_loss: 0.2719\n",
            "Epoch 34/40\n",
            "500/500 [==============================] - 393s 786ms/step - loss: 0.2574 - val_loss: 0.2746\n",
            "Epoch 35/40\n",
            "500/500 [==============================] - 393s 786ms/step - loss: 0.2565 - val_loss: 0.2689\n",
            "Epoch 36/40\n",
            "500/500 [==============================] - 393s 787ms/step - loss: 0.2559 - val_loss: 0.2762\n",
            "Epoch 37/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2535 - val_loss: 0.2702\n",
            "Epoch 38/40\n",
            "500/500 [==============================] - 395s 790ms/step - loss: 0.2548 - val_loss: 0.2724\n",
            "Epoch 39/40\n",
            "500/500 [==============================] - 394s 788ms/step - loss: 0.2532 - val_loss: 0.2715\n",
            "Epoch 40/40\n",
            "500/500 [==============================] - 391s 783ms/step - loss: 0.2532 - val_loss: 0.2771\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IsJ_vXbdBoL9",
        "colab_type": "code",
        "outputId": "51efaf2e-f670-4da5-e6ec-e22517931fcf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "plotLoss(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxGTJwhhjTEz/D4CTP0pzXcp8AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "14tHNp1IBpgE",
        "colab_type": "text"
      },
      "source": [
        "We see that adding an additional RNN layer does improve the results, though not significantly. \n",
        "\n",
        "### Bidirectional RNNs\n",
        "\n",
        "A bidirectional RNN is  formed of two regular RNNs, each of which processes the input sequence in one direction (chronologically and anti-chronologically). The represenations are then merged.\n",
        "\n",
        "To start: does it make a difference in performance if we go anti-chronologically when processing our sequence data? "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3tLP9wGxCT6y",
        "colab_type": "code",
        "outputId": "b1830e6c-0896-47f2-d452-51194ef94b86",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1617
        }
      },
      "source": [
        "# Generator yielding timeseries samples and their targets\n",
        "\n",
        "def time_reversed_generator(data, lookback, delay, min_index, max_index, shuffle=False, batch_size=128, step=6):\n",
        "  \"\"\"\n",
        "  Parameters\n",
        "  ----------\n",
        "  data : 2d numpy array\n",
        "      The array of floating-point timeseries data, normalized.\n",
        "  lookback : int\n",
        "      How many timesteps back the input data should go.\n",
        "  delay: int\n",
        "      How many timesteps in the future the target should be.\n",
        "  min_index: int\n",
        "      The lower bound of timesteps to draw from (helps keep validation/test data separate) \n",
        "  max_index: int\n",
        "      The upper bound of timesteps to draw from (helps keep validation/test data separate)\n",
        "  shuffle: bool\n",
        "      Whether to shuffle the samples or draw them in chronological order\n",
        "  batch_size: int\n",
        "      The number of samples per batch.\n",
        "  step: int\n",
        "      The period, in timesteps, at which you sample data.\n",
        "  \"\"\"\n",
        "  if max_index is None:\n",
        "    max_index = len(data) - delay - 1\n",
        "  i = min_index + lookback\n",
        "  while True:\n",
        "    if shuffle:\n",
        "      rows = np.random.randint(min_index+lookback, max_index, size=batch_size)\n",
        "    else:\n",
        "      if i+batch_size >= max_index:\n",
        "        i = min_index + lookback\n",
        "      rows = np.arange(i, min(i+batch_size, max_index))\n",
        "      i += len(rows)\n",
        "      \n",
        "    samples = np.zeros((len(rows), lookback // step, data.shape[-1]))\n",
        "    targets = np.zeros((len(rows),))\n",
        "    for j, row in enumerate(rows):\n",
        "      indices = range(rows[j] - lookback, rows[j], step)\n",
        "      samples[j] = data[indices]\n",
        "      targets[j] = data[rows[j] + delay][1]  # temperature is the second column\n",
        "    yield samples[:,::-1,:], targets\n",
        "    \n",
        "\n",
        "# Create a training, validation, and test generator looking at different segments of the data\n",
        "lookback = 1440\n",
        "step = 6\n",
        "delay = 144\n",
        "batch_size = 128\n",
        "\n",
        "time_reversed_train_gen = time_reversed_generator(float_data,\n",
        "                      lookback=lookback,\n",
        "                      delay=delay,\n",
        "                      min_index=0,\n",
        "                      max_index=200000,\n",
        "                      shuffle=True,\n",
        "                      step=step,\n",
        "                      batch_size=batch_size\n",
        "                     )\n",
        "time_reversed_val_gen = time_reversed_generator(float_data,\n",
        "                    lookback=lookback,\n",
        "                    delay=delay,\n",
        "                    min_index=200001,\n",
        "                    max_index=300000,\n",
        "                    step=step,\n",
        "                    batch_size=batch_size\n",
        "                   )\n",
        "\n",
        "# How many steps to draw from val_gen in order to see the entire validation set\n",
        "val_steps = (300000 - 200001 - lookback) // batch_size\n",
        "# As above, for the test set\n",
        "test_steps = (len(float_data) - 300001 - lookback) // batch_size\n",
        "    \n",
        "    \n",
        "\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.GRU(32,\n",
        "                           dropout=0.2,\n",
        "                           recurrent_dropout=0.2,\n",
        "                           input_shape=(None, float_data.shape[-1])))\n",
        "model.add(keras.layers.Dense(1))\n",
        "\n",
        "model.compile(optimizer=keras.optimizers.RMSprop(), loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(time_reversed_train_gen,\n",
        "                             steps_per_epoch=500,\n",
        "                             epochs=40,  # Twice as many epochs, as dropout makes convergence slower\n",
        "                             validation_data=time_reversed_val_gen,\n",
        "                             validation_steps=val_steps\n",
        "                             )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0620 12:35:16.971660 140290881169280 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "W0620 12:35:17.645499 140290881169280 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/40\n",
            "500/500 [==============================] - 253s 507ms/step - loss: 0.5036 - val_loss: 0.4801\n",
            "Epoch 2/40\n",
            "500/500 [==============================] - 253s 506ms/step - loss: 0.4882 - val_loss: 0.4765\n",
            "Epoch 3/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.4749 - val_loss: 0.4651\n",
            "Epoch 4/40\n",
            "500/500 [==============================] - 251s 501ms/step - loss: 0.4542 - val_loss: 0.4431\n",
            "Epoch 5/40\n",
            "500/500 [==============================] - 252s 503ms/step - loss: 0.4190 - val_loss: 0.4169\n",
            "Epoch 6/40\n",
            "500/500 [==============================] - 248s 497ms/step - loss: 0.3956 - val_loss: 0.4041\n",
            "Epoch 7/40\n",
            "500/500 [==============================] - 250s 500ms/step - loss: 0.3776 - val_loss: 0.3662\n",
            "Epoch 8/40\n",
            "500/500 [==============================] - 250s 500ms/step - loss: 0.3647 - val_loss: 0.3483\n",
            "Epoch 9/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.3549 - val_loss: 0.3429\n",
            "Epoch 10/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.3435 - val_loss: 0.3336\n",
            "Epoch 11/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.3356 - val_loss: 0.3258\n",
            "Epoch 12/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.3317 - val_loss: 0.3282\n",
            "Epoch 13/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.3248 - val_loss: 0.3056\n",
            "Epoch 14/40\n",
            "500/500 [==============================] - 251s 501ms/step - loss: 0.3139 - val_loss: 0.2968\n",
            "Epoch 15/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.3067 - val_loss: 0.2938\n",
            "Epoch 16/40\n",
            "500/500 [==============================] - 252s 503ms/step - loss: 0.3011 - val_loss: 0.2941\n",
            "Epoch 17/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2962 - val_loss: 0.2869\n",
            "Epoch 18/40\n",
            "500/500 [==============================] - 250s 501ms/step - loss: 0.2924 - val_loss: 0.2878\n",
            "Epoch 19/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2888 - val_loss: 0.2898\n",
            "Epoch 20/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2880 - val_loss: 0.2882\n",
            "Epoch 21/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2860 - val_loss: 0.2876\n",
            "Epoch 22/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.2845 - val_loss: 0.2888\n",
            "Epoch 23/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2837 - val_loss: 0.2936\n",
            "Epoch 24/40\n",
            "500/500 [==============================] - 251s 502ms/step - loss: 0.2837 - val_loss: 0.2908\n",
            "Epoch 25/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.2815 - val_loss: 0.2934\n",
            "Epoch 26/40\n",
            "500/500 [==============================] - 252s 503ms/step - loss: 0.2798 - val_loss: 0.2905\n",
            "Epoch 27/40\n",
            "500/500 [==============================] - 251s 503ms/step - loss: 0.2796 - val_loss: 0.2913\n",
            "Epoch 28/40\n",
            "500/500 [==============================] - 250s 500ms/step - loss: 0.2780 - val_loss: 0.2965\n",
            "Epoch 29/40\n",
            "500/500 [==============================] - 250s 500ms/step - loss: 0.2785 - val_loss: 0.2988\n",
            "Epoch 30/40\n",
            "500/500 [==============================] - 252s 504ms/step - loss: 0.2773 - val_loss: 0.2914\n",
            "Epoch 31/40\n",
            "500/500 [==============================] - 252s 504ms/step - loss: 0.2753 - val_loss: 0.2934\n",
            "Epoch 32/40\n",
            "500/500 [==============================] - 250s 501ms/step - loss: 0.2758 - val_loss: 0.2930\n",
            "Epoch 33/40\n",
            "500/500 [==============================] - 249s 498ms/step - loss: 0.2754 - val_loss: 0.2982\n",
            "Epoch 34/40\n",
            "500/500 [==============================] - 249s 498ms/step - loss: 0.2726 - val_loss: 0.2974\n",
            "Epoch 35/40\n",
            "500/500 [==============================] - 250s 500ms/step - loss: 0.2722 - val_loss: 0.2995\n",
            "Epoch 36/40\n",
            "500/500 [==============================] - 249s 497ms/step - loss: 0.2738 - val_loss: 0.2987\n",
            "Epoch 37/40\n",
            "500/500 [==============================] - 248s 495ms/step - loss: 0.2718 - val_loss: 0.3037\n",
            "Epoch 38/40\n",
            "500/500 [==============================] - 249s 499ms/step - loss: 0.2712 - val_loss: 0.2949\n",
            "Epoch 39/40\n",
            "500/500 [==============================] - 246s 493ms/step - loss: 0.2724 - val_loss: 0.2978\n",
            "Epoch 40/40\n",
            "500/500 [==============================] - 249s 498ms/step - loss: 0.2724 - val_loss: 0.2988\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LLBzxHJO2p9A",
        "colab_type": "code",
        "outputId": "178cda1b-de88-47cd-c586-9f6e78af8b38",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "plotLoss(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GmfeU4BC4mCV",
        "colab_type": "text"
      },
      "source": [
        "We see that even  our second-best model (GRU with dropout) underperforms the common-sense baseline, badly - this makes sense given our dataset (and perhaps wouldn't be true for e.g. NLP) .\n",
        "\n",
        "Let's try the same on our IMDB dataset:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_uytheVi462l",
        "colab_type": "code",
        "outputId": "5cb66055-0477-4cd9-feed-33922b9ab910",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 439
        }
      },
      "source": [
        "max_features=10000\n",
        "max_length = 500\n",
        "\n",
        "# Overwrite np.load to allow keras to work with the new Numpy defaults\n",
        "old = np.load\n",
        "np.load = lambda *a,**k: old(*a,**k,allow_pickle=True)\n",
        "(X_train, y_train), (X_test, y_test) = keras.datasets.imdb.load_data(num_words=max_features)\n",
        "np.load = old\n",
        "del(old)\n",
        "\n",
        "# Reverse the sequences\n",
        "X_train = [x[::-1] for x in X_train]\n",
        "X_test = [x[::-1] for x in X_test]\n",
        "\n",
        "# Turns the list of integers into a 2D integer tensor\n",
        "X_train = keras.preprocessing.sequence.pad_sequences(X_train, maxlen=max_length)\n",
        "X_test = keras.preprocessing.sequence.pad_sequences(X_test, maxlen=max_length)\n",
        "\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_features, 128))\n",
        "model.add(keras.layers.LSTM(32))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "\n",
        "history = model.fit(X_train, y_train,\n",
        "                    epochs=10,\n",
        "                    batch_size=128,\n",
        "                    validation_split=0.2\n",
        "                   )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://s3.amazonaws.com/text-datasets/imdb.npz\n",
            "17465344/17464789 [==============================] - 1s 0us/step\n",
            "Train on 20000 samples, validate on 5000 samples\n",
            "Epoch 1/10\n",
            "20000/20000 [==============================] - 145s 7ms/step - loss: 0.4912 - acc: 0.7679 - val_loss: 0.4331 - val_acc: 0.8350\n",
            "Epoch 2/10\n",
            "20000/20000 [==============================] - 143s 7ms/step - loss: 0.3171 - acc: 0.8772 - val_loss: 0.3231 - val_acc: 0.8686\n",
            "Epoch 3/10\n",
            "20000/20000 [==============================] - 144s 7ms/step - loss: 0.2633 - acc: 0.9015 - val_loss: 0.3353 - val_acc: 0.8582\n",
            "Epoch 4/10\n",
            "20000/20000 [==============================] - 143s 7ms/step - loss: 0.2243 - acc: 0.9170 - val_loss: 0.4565 - val_acc: 0.8518\n",
            "Epoch 5/10\n",
            "20000/20000 [==============================] - 143s 7ms/step - loss: 0.1947 - acc: 0.9310 - val_loss: 0.3515 - val_acc: 0.8754\n",
            "Epoch 6/10\n",
            "20000/20000 [==============================] - 143s 7ms/step - loss: 0.1784 - acc: 0.9376 - val_loss: 0.3469 - val_acc: 0.8748\n",
            "Epoch 7/10\n",
            "20000/20000 [==============================] - 144s 7ms/step - loss: 0.1623 - acc: 0.9437 - val_loss: 0.4216 - val_acc: 0.8728\n",
            "Epoch 8/10\n",
            "20000/20000 [==============================] - 144s 7ms/step - loss: 0.1397 - acc: 0.9525 - val_loss: 0.7025 - val_acc: 0.7918\n",
            "Epoch 9/10\n",
            "20000/20000 [==============================] - 143s 7ms/step - loss: 0.1360 - acc: 0.9549 - val_loss: 0.3749 - val_acc: 0.8544\n",
            "Epoch 10/10\n",
            "20000/20000 [==============================] - 142s 7ms/step - loss: 0.1198 - acc: 0.9605 - val_loss: 0.4947 - val_acc: 0.8380\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4nQ0jCWketJw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Plot the loss and accuracy of the model over the training and validation data\n",
        "def plotLossAndAccuracy(history):\n",
        "  history_dict = history.history\n",
        "  acc = history_dict[\"acc\"]\n",
        "  val_acc = history_dict[\"val_acc\"]\n",
        "  loss = history_dict[\"loss\"]\n",
        "  val_loss = history_dict[\"val_loss\"]\n",
        "  epochs = range(1, len(loss) +1)\n",
        "\n",
        "  plt.plot(epochs, acc, \"bo\", label=\"Training Accuracy\")\n",
        "  plt.plot(epochs, val_acc, \"b\", label=\"Validation Accuracy\")\n",
        "  plt.title(\"Training and validation accuracy\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Accuracy\")\n",
        "  plt.legend()\n",
        "  plt.show()\n",
        "\n",
        "  plt.plot(epochs, loss, \"bo\", label=\"Training Loss\")\n",
        "  plt.plot(epochs, val_loss, \"b\", label=\"Validation Loss\")\n",
        "  plt.title(\"Training and validation loss\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Loss\")\n",
        "  plt.legend()\n",
        "  plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oty_TBIJ6Nd8",
        "colab_type": "code",
        "outputId": "3cb48cc8-f425-4f17-9ae3-30946a5e8b7c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "plotLossAndAccuracy(history) # Accuracy was about 0.88 for chronological LSTM on this data"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRGKMMSYslkiMMcaExRKJMcaYsPw/fU7SrOeFQaUAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1cUlAo9eeumi",
        "colab_type": "text"
      },
      "source": [
        "So for the text-analysis problem, the performance of the forwards and backwards LSTMs are roughly equivalent.\n",
        "\n",
        "Let's try a `Bidirectional` layer on the IMDB data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BoGRq4kfxtky",
        "colab_type": "code",
        "outputId": "821c5924-698c-4f6c-b08f-062805b65a98",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 403
        }
      },
      "source": [
        "max_features=10000\n",
        "max_length = 500\n",
        "\n",
        "# Overwrite np.load to allow keras to work with the new Numpy defaults\n",
        "old = np.load\n",
        "np.load = lambda *a,**k: old(*a,**k,allow_pickle=True)\n",
        "(X_train, y_train), (X_test, y_test) = keras.datasets.imdb.load_data(num_words=max_features)\n",
        "np.load = old\n",
        "del(old)\n",
        "\n",
        "\n",
        "# Turns the list of integers into a 2D integer tensor\n",
        "X_train = keras.preprocessing.sequence.pad_sequences(X_train, maxlen=max_length)\n",
        "X_test = keras.preprocessing.sequence.pad_sequences(X_test, maxlen=max_length)\n",
        "\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Embedding(max_features, 128))\n",
        "model.add(keras.layers.Bidirectional(keras.layers.LSTM(32)))\n",
        "model.add(keras.layers.Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "\n",
        "history = model.fit(X_train, y_train,\n",
        "                    epochs=10,\n",
        "                    batch_size=128,\n",
        "                    validation_split=0.2\n",
        "                   )"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 20000 samples, validate on 5000 samples\n",
            "Epoch 1/10\n",
            "20000/20000 [==============================] - 287s 14ms/step - loss: 0.4773 - acc: 0.7742 - val_loss: 0.3572 - val_acc: 0.8504\n",
            "Epoch 2/10\n",
            "20000/20000 [==============================] - 288s 14ms/step - loss: 0.2906 - acc: 0.8873 - val_loss: 0.2961 - val_acc: 0.8814\n",
            "Epoch 3/10\n",
            "20000/20000 [==============================] - 287s 14ms/step - loss: 0.2289 - acc: 0.9146 - val_loss: 0.2963 - val_acc: 0.8880\n",
            "Epoch 4/10\n",
            "20000/20000 [==============================] - 294s 15ms/step - loss: 0.1985 - acc: 0.9277 - val_loss: 0.3047 - val_acc: 0.8848\n",
            "Epoch 5/10\n",
            "20000/20000 [==============================] - 275s 14ms/step - loss: 0.1700 - acc: 0.9380 - val_loss: 0.3002 - val_acc: 0.8872\n",
            "Epoch 6/10\n",
            "20000/20000 [==============================] - 286s 14ms/step - loss: 0.1500 - acc: 0.9453 - val_loss: 0.3692 - val_acc: 0.8808\n",
            "Epoch 7/10\n",
            "20000/20000 [==============================] - 293s 15ms/step - loss: 0.1304 - acc: 0.9541 - val_loss: 0.3715 - val_acc: 0.8734\n",
            "Epoch 8/10\n",
            "20000/20000 [==============================] - 288s 14ms/step - loss: 0.1219 - acc: 0.9573 - val_loss: 0.3491 - val_acc: 0.8850\n",
            "Epoch 9/10\n",
            "20000/20000 [==============================] - 288s 14ms/step - loss: 0.1107 - acc: 0.9607 - val_loss: 0.4050 - val_acc: 0.8726\n",
            "Epoch 10/10\n",
            "20000/20000 [==============================] - 286s 14ms/step - loss: 0.1013 - acc: 0.9645 - val_loss: 0.4376 - val_acc: 0.8760\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nIseDxDRx_nS",
        "colab_type": "code",
        "outputId": "e45df4ed-636c-49e7-f6e3-41e1ba0ad2b3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "plotLossAndAccuracy(history)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZPcAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p5PbccOmyEvt",
        "colab_type": "text"
      },
      "source": [
        "We get an accuracy of ~0.89%, slightly better than the regular LSTM. We also overfit (since we now have twice as many parameters). Let's try the same approach on the temperature-prediction task:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kYqvjsdByUUR",
        "colab_type": "code",
        "outputId": "6cbacd5d-414d-4474-fa3d-7574d32d2588",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 568
        }
      },
      "source": [
        "\n",
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Bidirectional(keras.layers.GRU(32),\n",
        "                                     input_shape=(None, float_data.shape[-1])))\n",
        "model.add(keras.layers.Dense(1))\n",
        "\n",
        "model.compile(optimizer=\"rmsprop\", loss=\"mae\")\n",
        "\n",
        "history = model.fit_generator(train_gen,\n",
        "                              steps_per_epoch=500,\n",
        "                              epochs=40,\n",
        "                              validation_data=val_gen,\n",
        "                              validation_steps=val_steps)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/40\n",
            "500/500 [==============================] - 418s 835ms/step - loss: 0.2948 - val_loss: 0.2808\n",
            "Epoch 2/40\n",
            "500/500 [==============================] - 419s 838ms/step - loss: 0.2753 - val_loss: 0.2717\n",
            "Epoch 3/40\n",
            "500/500 [==============================] - 414s 827ms/step - loss: 0.2670 - val_loss: 0.2634\n",
            "Epoch 4/40\n",
            "500/500 [==============================] - 418s 837ms/step - loss: 0.2614 - val_loss: 0.2731\n",
            "Epoch 5/40\n",
            "500/500 [==============================] - 415s 830ms/step - loss: 0.2558 - val_loss: 0.2758\n",
            "Epoch 6/40\n",
            "500/500 [==============================] - 415s 829ms/step - loss: 0.2525 - val_loss: 0.2746\n",
            "Epoch 7/40\n",
            "500/500 [==============================] - 414s 828ms/step - loss: 0.2465 - val_loss: 0.2790\n",
            "Epoch 8/40\n",
            "500/500 [==============================] - 417s 835ms/step - loss: 0.2408 - val_loss: 0.2822\n",
            "Epoch 9/40\n",
            "500/500 [==============================] - 411s 822ms/step - loss: 0.2330 - val_loss: 0.2771\n",
            "Epoch 10/40\n",
            "500/500 [==============================] - 414s 829ms/step - loss: 0.2284 - val_loss: 0.2846\n",
            "Epoch 11/40\n",
            "500/500 [==============================] - 415s 831ms/step - loss: 0.2226 - val_loss: 0.2896\n",
            "Epoch 12/40\n",
            "500/500 [==============================] - 413s 825ms/step - loss: 0.2176 - val_loss: 0.2942\n",
            "Epoch 13/40\n",
            "500/500 [==============================] - 419s 839ms/step - loss: 0.2124 - val_loss: 0.2940\n",
            "Epoch 14/40\n",
            "500/500 [==============================] - 419s 838ms/step - loss: 0.2073 - val_loss: 0.3016\n",
            "Epoch 15/40\n",
            "441/500 [=========================>....] - ETA: 38s - loss: 0.2028"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "riHgPBW3y1xD",
        "colab_type": "text"
      },
      "source": [
        "This timed out, but it seems like it performs about as well as the chronological GRU (as you might expect, since the antichron GRU is bad)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cV9UFIJeWfBK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}