2-3-9-AssignmentQuestion.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iiVBUpuHXEtw"
      },
      "source": [
        "# Create a CNN to classify Cifar-10\n",
        "\n",
        "Learn about Cifar-10 here: https://www.cs.toronto.edu/~kriz/cifar.html\n",
        "\n",
        "In class you saw how to build a Convolutional Neural Network that classified Fashion MNIST. Take what you learned to build a CNN that recognizes the 10 classes of CIFAR. It will be a similar network, but there are some key differences you'll need to take into account.\n",
        "\n",
        "First, while MNIST were 28x28 monochome images (1 color channel), CIFAR are 32x32 color images (3 color channels).\n",
        "\n",
        "Second, MNIST images are simple, containing just the object, centered in the image, with no background. CIFAR ones can have the object with a background -- for example airplanes might have a cloudy sky behind them! As such you should expect your accuracy to be a bit lower.\n",
        "\n",
        "We start by setting up the problem for you."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "6OQ_tVTaU3oo",
        "outputId": "35f78e4e-3605-4b10-9ad9-a974b2c7a255",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "\u001b[1m170498071/170498071\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 0us/step\n"
          ]
        }
      ],
      "source": [
        "import tensorflow as tf\n",
        "\n",
        "from tensorflow.keras import datasets, layers, models\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()\n",
        "\n",
        "# Normalize pixel values to be between 0 and 1\n",
        "train_images = train_images / 255.0\n",
        "test_images = test_images / 255.0"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CWc6D_PRU3MS"
      },
      "source": [
        "We then definte some of the model for you but leave most of it for you to fill in!\n",
        "\n",
        "*A hint: your model may want to learn some high level features and then classify them.*"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "law2hNzdVA16"
      },
      "outputs": [],
      "source": [
        "FIRST_LAYER = tf.keras.layers.Conv2D(64, (3,3), activation='relu', input_shape=(32,32,3))\n",
        "HIDDEN_LAYER_TYPE_1 = tf.keras.layers.MaxPooling2D((2,2))\n",
        "HIDDEN_LAYER_TYPE_2 = tf.keras.layers.Conv2D(64, (3,3), activation='relu')\n",
        "HIDDEN_LAYER_TYPE_3 = tf.keras.layers.MaxPooling2D((2,2))\n",
        "HIDDEN_LAYER_TYPE_4 = tf.keras.layers.Dense(128, activation='relu')\n",
        "HIDDEN_LAYER_TYPE_5 = tf.keras.layers.Dense(64, activation='relu')\n",
        "LAST_LAYER = tf.keras.layers.Dense(10, activation='softmax')\n",
        "\n",
        "model = models.Sequential([\n",
        "       FIRST_LAYER,\n",
        "       HIDDEN_LAYER_TYPE_1,\n",
        "       HIDDEN_LAYER_TYPE_2,\n",
        "       HIDDEN_LAYER_TYPE_3,\n",
        "       layers.Flatten(),\n",
        "       HIDDEN_LAYER_TYPE_4,\n",
        "       HIDDEN_LAYER_TYPE_5,\n",
        "       LAST_LAYER,\n",
        "])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CFvunU4jU3MT"
      },
      "source": [
        "You then need to define loss function. And you can then train your model. Once training is done you'll see a plot of training and validation accuracy. You'll know you have a reasonable model with a reasonable loss funciton if your final training accuracy ends up in the 70s (or possibly higher).\n",
        "\n",
        "*A hint: your model may want to learn different categories.*"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "Dxq2JU7dVIEX",
        "outputId": "a7f4c067-b2dc-46d3-b857-db344b8754fd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 4ms/step - accuracy: 0.2139 - loss: 2.1327 - val_accuracy: 0.3817 - val_loss: 1.7314\n",
            "Epoch 2/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.3976 - loss: 1.6755 - val_accuracy: 0.4703 - val_loss: 1.4622\n",
            "Epoch 3/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.4784 - loss: 1.4520 - val_accuracy: 0.4416 - val_loss: 1.5173\n",
            "Epoch 4/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.5201 - loss: 1.3465 - val_accuracy: 0.5325 - val_loss: 1.2932\n",
            "Epoch 5/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.5599 - loss: 1.2501 - val_accuracy: 0.5547 - val_loss: 1.2431\n",
            "Epoch 6/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.5845 - loss: 1.1682 - val_accuracy: 0.5887 - val_loss: 1.1683\n",
            "Epoch 7/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.6145 - loss: 1.0938 - val_accuracy: 0.5934 - val_loss: 1.1488\n",
            "Epoch 8/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.6375 - loss: 1.0361 - val_accuracy: 0.6260 - val_loss: 1.0609\n",
            "Epoch 9/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.6642 - loss: 0.9598 - val_accuracy: 0.6498 - val_loss: 1.0056\n",
            "Epoch 10/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.6847 - loss: 0.9130 - val_accuracy: 0.6489 - val_loss: 1.0058\n",
            "Epoch 11/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.6965 - loss: 0.8679 - val_accuracy: 0.6376 - val_loss: 1.0681\n",
            "Epoch 12/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.7131 - loss: 0.8202 - val_accuracy: 0.6460 - val_loss: 1.0210\n",
            "Epoch 13/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.7291 - loss: 0.7731 - val_accuracy: 0.6733 - val_loss: 0.9498\n",
            "Epoch 14/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 3ms/step - accuracy: 0.7454 - loss: 0.7359 - val_accuracy: 0.6768 - val_loss: 0.9543\n",
            "Epoch 15/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.7520 - loss: 0.7127 - val_accuracy: 0.6730 - val_loss: 0.9760\n",
            "Epoch 16/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.7715 - loss: 0.6607 - val_accuracy: 0.6623 - val_loss: 1.0291\n",
            "Epoch 17/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.7836 - loss: 0.6257 - val_accuracy: 0.6903 - val_loss: 0.9242\n",
            "Epoch 18/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.7933 - loss: 0.5908 - val_accuracy: 0.6668 - val_loss: 1.0347\n",
            "Epoch 19/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.8089 - loss: 0.5511 - val_accuracy: 0.6900 - val_loss: 0.9632\n",
            "Epoch 20/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.8253 - loss: 0.5112 - val_accuracy: 0.6832 - val_loss: 1.0051\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "LOSS = 'sparse_categorical_crossentropy'\n",
        "NUM_EPOCHS = 20\n",
        "\n",
        "# Compile the model\n",
        "model.compile(optimizer='sgd',\n",
        "              loss=LOSS,\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "# Fit the model\n",
        "history = model.fit(train_images, train_labels, epochs=NUM_EPOCHS,\n",
        "                    validation_data=(test_images, test_labels))\n",
        "\n",
        "# summarize history for accuracy\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.xlim([0,NUM_EPOCHS])\n",
        "plt.ylim([0.4,1.0])\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LCKMjR-9U3MT"
      },
      "source": [
        "Finally, pick a better optimizer. And re-train your model. You'll know you have a reasonable model with a reasonable loss funciton and optimizer if your final training accuracy ends up in the 80s (or possibly higher).\n",
        "\n",
        "*A hint: your model may want to learn adaptively.*"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "ERAdd2qAU3MT",
        "outputId": "fce48c6b-fe49-4230-9474-df29a5a8b7e2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 4ms/step - accuracy: 0.6948 - loss: 0.8705 - val_accuracy: 0.6801 - val_loss: 0.9493\n",
            "Epoch 2/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.7597 - loss: 0.6955 - val_accuracy: 0.6881 - val_loss: 0.9440\n",
            "Epoch 3/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.8056 - loss: 0.5524 - val_accuracy: 0.6983 - val_loss: 0.9545\n",
            "Epoch 4/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.8456 - loss: 0.4360 - val_accuracy: 0.6904 - val_loss: 1.0160\n",
            "Epoch 5/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.8817 - loss: 0.3354 - val_accuracy: 0.6917 - val_loss: 1.1234\n",
            "Epoch 6/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.9080 - loss: 0.2633 - val_accuracy: 0.7036 - val_loss: 1.1274\n",
            "Epoch 7/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.9264 - loss: 0.2074 - val_accuracy: 0.7027 - val_loss: 1.2385\n",
            "Epoch 8/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 5ms/step - accuracy: 0.9406 - loss: 0.1706 - val_accuracy: 0.7010 - val_loss: 1.3859\n",
            "Epoch 9/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9476 - loss: 0.1478 - val_accuracy: 0.6924 - val_loss: 1.5503\n",
            "Epoch 10/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9543 - loss: 0.1301 - val_accuracy: 0.6855 - val_loss: 1.6211\n",
            "Epoch 11/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9632 - loss: 0.1077 - val_accuracy: 0.6911 - val_loss: 1.7427\n",
            "Epoch 12/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9611 - loss: 0.1097 - val_accuracy: 0.6930 - val_loss: 1.7220\n",
            "Epoch 13/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.9668 - loss: 0.0976 - val_accuracy: 0.6975 - val_loss: 1.8193\n",
            "Epoch 14/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9698 - loss: 0.0884 - val_accuracy: 0.6918 - val_loss: 1.8530\n",
            "Epoch 15/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9713 - loss: 0.0861 - val_accuracy: 0.6897 - val_loss: 1.8702\n",
            "Epoch 16/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9744 - loss: 0.0742 - val_accuracy: 0.6982 - val_loss: 2.0245\n",
            "Epoch 17/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9730 - loss: 0.0801 - val_accuracy: 0.6884 - val_loss: 2.0681\n",
            "Epoch 18/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 5ms/step - accuracy: 0.9771 - loss: 0.0694 - val_accuracy: 0.6899 - val_loss: 2.1620\n",
            "Epoch 19/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 3ms/step - accuracy: 0.9782 - loss: 0.0662 - val_accuracy: 0.6919 - val_loss: 2.2261\n",
            "Epoch 20/20\n",
            "\u001b[1m1563/1563\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9787 - loss: 0.0665 - val_accuracy: 0.6904 - val_loss: 2.2174\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "OPTIMIZER = tf.keras.optimizers.Adam()\n",
        "\n",
        "# Compile the model\n",
        "model.compile(optimizer=OPTIMIZER,\n",
        "              loss=LOSS,\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "# Fit the model\n",
        "history = model.fit(train_images, train_labels, epochs=NUM_EPOCHS,\n",
        "                    validation_data=(test_images, test_labels))\n",
        "\n",
        "# summarize history for accuracy\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.xlim([0,NUM_EPOCHS])\n",
        "plt.ylim([0.4,1.0])\n",
        "plt.show()"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "2-3-9-Question.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.9"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}