A1_BSc.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cIg6enKVhhi4"
      },
      "source": [
        "\n",
        "\n",
        "## **COMP6685 Deep Learning Coursework A1**\n",
        "\n",
        "\n",
        "Individual (25% of total mark)\n",
        "\n",
        "\n",
        "**TASK:**\tYou are required to develop a phyton code with appropriate comments and answer questions.\n",
        "\n",
        "**Description**: Create a code using this temlate to train a Convolutional Neural Network (CNN) on the fashion MNIST dataset available at https://keras.io/api/datasets/fashion_mnist/ . \n",
        "\n",
        "Fashion MNIST is a dataset of 60,000 28x28 grayscale images of 10 fashion categories, along with a test set of 10,000 images.\n",
        "\n",
        "The dataset should be imported in the code and one sample image should be visualised before applying the model.\n",
        "\n",
        "Define a CNN and comment the chosen parameters of the network. Apply a regularization method (L1, L2 or L1L2). Divide the dataset into training, validation and test set. Obtain the accuracy on the validation set and plot the final results using the data from the test set. Comment your lines of code appropriately to explain your solution.\n",
        "\n",
        "Enhance the model's performance to obtain the best or optimal validation accuracy. Further questions about final remarks on the results will be answered on the markdown defined in the template.\n",
        "\n",
        "---\n",
        "---\n",
        "\n",
        "Note: This is only a template. You can add more code/text cells if necessary."
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {
        "id": "qysRQbvOm1c4"
      },
      "source": [
        "Import the dataset and divide it into training, validation and test sets. Explain how you obtained the validation set. How did you choose the size of the validation set? **(10 marks)**\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 220,
      "metadata": {
        "id": "tV9PO1GBl5_6"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "# Load the Fashion MNIST dataset using Keras\n",
        "fashion_mnist = keras.datasets.fashion_mnist\n",
        "\n",
        "# Split the dataset into training and test sets\n",
        "(train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()\n",
        "\n",
        "# Determine the split index for training and validation data\n",
        "split_index = int(len(train_images) * 0.7)\n",
        "\n",
        "# Split the training data into training and validation sets\n",
        "validation_images = train_images[split_index:]\n",
        "validation_labels = train_labels[split_index:]\n",
        "train_images = train_images[:split_index]\n",
        "train_labels = train_labels[:split_index]\n",
        "\n"
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "*your answer here*\n",
        "\n",
        "Firstly I split the dataset into training and test sets. I then defined a point at which to split the training data in order to create a validation set by using 20% of the training set. I chose to use 20% as this is a common practice and as the fashionMNIST dataset is large I believe this will give me enough data to use for evaluating the performance of my model. \n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 221,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Each 2D image consists of 28x28 values/pixels, which needs to be reshaped in a vector of 784 pixels\n",
        "train_images = train_images.reshape((train_images.shape[0], 784))\n",
        "test_images = test_images.reshape((test_images.shape[0], 784))\n",
        "validation_images = validation_images.reshape((validation_images.shape[0], 784))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 222,
      "metadata": {},
      "outputs": [],
      "source": [
        "# By dividing the pixel values by 255, all the values are scaled to be between 0 and 1. \n",
        "# This normalization step helps to ensure that the model is less sensitive to the scale of the input data\n",
        "# Which can lead to better performance and faster convergence during training.\n",
        "train_images = train_images / 255.0\n",
        "test_images = test_images / 255.0\n",
        "validation_images = validation_images / 255.0\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2-xHphsvmG7F"
      },
      "source": [
        "Visualise a random sample image of the dataset. **(10 marks)**\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 223,
      "metadata": {
        "id": "bTTrIN_nnDWK"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "#select a random image from the training set and display it\n",
        "random_index = np.random.randint(0, train_images.shape[0])\n",
        "#reshape the image to 28x28 pixels so it can be displayed\n",
        "image = train_images[random_index].reshape((28, 28))\n",
        "plt.imshow(image, cmap='gray')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lMhGFcimoUwz"
      },
      "source": [
        "Define your CNN model. Specify the network and training parameters and comment them. **(10 marks)**\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 224,
      "metadata": {},
      "outputs": [],
      "source": [
        "from keras.utils import to_categorical\n",
        "\n",
        "# Convert labels to one-hot encoded format\n",
        "train_labels = to_categorical(train_labels)\n",
        "validation_labels = to_categorical(validation_labels)\n",
        "test_labels = to_categorical(test_labels)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 225,
      "metadata": {
        "id": "urupEGOFooUz"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential_19\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense_57 (Dense)            (None, 128)               100480    \n",
            "                                                                 \n",
            " activation_55 (Activation)  (None, 128)               0         \n",
            "                                                                 \n",
            " dropout_38 (Dropout)        (None, 128)               0         \n",
            "                                                                 \n",
            " dense_58 (Dense)            (None, 128)               16512     \n",
            "                                                                 \n",
            " activation_56 (Activation)  (None, 128)               0         \n",
            "                                                                 \n",
            " dropout_39 (Dropout)        (None, 128)               0         \n",
            "                                                                 \n",
            " dense_59 (Dense)            (None, 10)                1290      \n",
            "                                                                 \n",
            " activation_57 (Activation)  (None, 10)                0         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 118,282\n",
            "Trainable params: 118,282\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "from keras.layers import Activation, Dropout\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense\n",
        "from tensorflow.keras import regularizers\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "\n",
        "OPTIMIZER = SGD(learning_rate=0.2) # Stochastic gradient descent optimiser\n",
        "N_EPOCH = 20 \n",
        "N_HIDDEN = 128\n",
        "\n",
        "model = Sequential()\n",
        "\n",
        "# Hidden layer 1 with 128 hidden units, ReLu activation function, and dropout rate of 0.2\n",
        "model.add(Dense(N_HIDDEN, kernel_regularizer=regularizers.L1L2(l1=1e-5, l2=1e-4), bias_regularizer=regularizers.L1(1e-5), activity_regularizer=regularizers.L2(1e-4), input_shape=(784,)))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.2))\n",
        "\n",
        "# Hidden layer 2 with 128 hidden units, ReLu activation function, and dropout rate of 0.2\n",
        "model.add(Dense(N_HIDDEN, kernel_regularizer=regularizers.L1L2(l1=1e-5, l2=1e-4), bias_regularizer=regularizers.L1(1e-5), activity_regularizer=regularizers.L2(1e-4)))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.2))\n",
        "\n",
        "# Output layer with 10 units and softmax activation\n",
        "model.add(Dense(10))\n",
        "model.add(Activation('softmax'))\n",
        "\n",
        "\n",
        "# Summary of the whole model\n",
        "model.summary()\n",
        "\n",
        "# model compilation\n",
        "model.compile(loss='binary_crossentropy', optimizer=OPTIMIZER, metrics=['accuracy'])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "csCYqn-9pDuJ"
      },
      "source": [
        "Train the CNN model. **(10 marks)**\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 226,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "329/329 [==============================] - 2s 4ms/step - loss: 0.3118 - accuracy: 0.5835 - val_loss: 0.2355 - val_accuracy: 0.7390\n",
            "Epoch 2/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.2325 - accuracy: 0.7401 - val_loss: 0.2077 - val_accuracy: 0.7833\n",
            "Epoch 3/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.2091 - accuracy: 0.7858 - val_loss: 0.1922 - val_accuracy: 0.8086\n",
            "Epoch 4/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1953 - accuracy: 0.8043 - val_loss: 0.1834 - val_accuracy: 0.8173\n",
            "Epoch 5/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1854 - accuracy: 0.8167 - val_loss: 0.1791 - val_accuracy: 0.8159\n",
            "Epoch 6/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1783 - accuracy: 0.8243 - val_loss: 0.1684 - val_accuracy: 0.8314\n",
            "Epoch 7/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1720 - accuracy: 0.8329 - val_loss: 0.1632 - val_accuracy: 0.8352\n",
            "Epoch 8/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1668 - accuracy: 0.8344 - val_loss: 0.1574 - val_accuracy: 0.8441\n",
            "Epoch 9/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1619 - accuracy: 0.8389 - val_loss: 0.1550 - val_accuracy: 0.8386\n",
            "Epoch 10/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1575 - accuracy: 0.8439 - val_loss: 0.1550 - val_accuracy: 0.8382\n",
            "Epoch 11/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1540 - accuracy: 0.8452 - val_loss: 0.1512 - val_accuracy: 0.8385\n",
            "Epoch 12/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1501 - accuracy: 0.8485 - val_loss: 0.1559 - val_accuracy: 0.8266\n",
            "Epoch 13/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1465 - accuracy: 0.8515 - val_loss: 0.1437 - val_accuracy: 0.8472\n",
            "Epoch 14/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1435 - accuracy: 0.8534 - val_loss: 0.1386 - val_accuracy: 0.8559\n",
            "Epoch 15/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1403 - accuracy: 0.8547 - val_loss: 0.1374 - val_accuracy: 0.8533\n",
            "Epoch 16/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1376 - accuracy: 0.8588 - val_loss: 0.1368 - val_accuracy: 0.8496\n",
            "Epoch 17/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1355 - accuracy: 0.8581 - val_loss: 0.1318 - val_accuracy: 0.8564\n",
            "Epoch 18/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1325 - accuracy: 0.8606 - val_loss: 0.1322 - val_accuracy: 0.8516\n",
            "Epoch 19/20\n",
            "329/329 [==============================] - 1s 3ms/step - loss: 0.1303 - accuracy: 0.8611 - val_loss: 0.1303 - val_accuracy: 0.8510\n",
            "Epoch 20/20\n",
            "329/329 [==============================] - 1s 4ms/step - loss: 0.1276 - accuracy: 0.8635 - val_loss: 0.1272 - val_accuracy: 0.8533\n"
          ]
        }
      ],
      "source": [
        "# Training the model\n",
        "history = model.fit(train_images, train_labels, validation_data=(validation_images, validation_labels), batch_size=128, epochs=N_EPOCH)\n"
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {
        "id": "FFMiFtDCpjMG"
      },
      "source": [
        "Evaluate your model. What is the best/highest validation accuracy your network achieved? How did you obtain this accuracy? **(10 marks)**\n",
        "\n",
        "l1l2 = 0.8828 acc 0.1137 loss\n",
        "\n",
        "no regularisation = 0.8773 acc 0.0599 loss\n",
        "\n",
        "L1 = 0.8761 acc 0.1057 loss\n",
        "\n",
        "L2 = 0.8778 acc 0.0924 loss \n"
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {
        "id": "4GvX1moIqqm0"
      },
      "source": [
        "(answer)\n",
        "The highest accuracy I achieved on the validation set was 0.8828. I obtained this accuracy by training the model for 20 epochs, using a batch size of 128, and using a learning rate of 0.2. I also used a dropout rate of 0.2 with a combination of L1(1e-5) and L2(1e-4) regularization.\n",
        "I chose to use a learning rate of 0.2 because it was the highest learning rate that did not cause the model to diverge. I chose to use a dropout rate of 0.2 because it is the highest rate that didnt cause the model to underfit the data, which means it may not be able to capture the important patterns in the data.\n",
        "\n",
        " I chose to use a combination of L1 and L2 regularization because it was the highest regularization rate that did not cause the model to underfit. I chose to train the model for 20 epochs because it was the highest number of epochs that did not cause the model to overfit. I chose to use binary cross entropy as the loss function becasue I had nomralised my data to be between 0 and 1 in order to minimise complexity of the model. I chose to use stocastic gradient descent as the optimiser becuase it gave me the best results when training the model."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 227,
      "metadata": {
        "id": "WnCeV6qP6xmT"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "\n",
            "Training score/loss: 0.12339572608470917\n",
            "Training accuracy: 0.8640714287757874\n",
            "\n",
            "Test score/loss: 0.13183729350566864\n",
            "Test accuracy: 0.8460000157356262\n",
            "\n",
            "Validation score/loss: 0.12724925577640533\n",
            "Validation accuracy: 0.8533333539962769\n"
          ]
        }
      ],
      "source": [
        "#test the network\n",
        "score = model.evaluate(train_images, train_labels, verbose=0)\n",
        "score1 = model.evaluate(test_images, test_labels, verbose=0)\n",
        "score2 = model.evaluate(validation_images, validation_labels, verbose=0)\n",
        "\n",
        "print(\"\\nTraining score/loss:\", score[0])\n",
        "print(\"Training accuracy:\", score[1])\n",
        "print(\"\\nTest score/loss:\", score1[0])\n",
        "print (\"Test accuracy:\", score1[1])\n",
        "print(\"\\nValidation score/loss:\", score2[0])\n",
        "print (\"Validation accuracy:\", score2[1])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S48e3kq_qCCi"
      },
      "source": [
        "Plot the final results on the test set and print the accuracy/loss on that set. **(10 marks)**\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 233,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 0s 1ms/step - loss: 0.1318 - accuracy: 0.8460\n",
            "Test accuracy: 0.8460000157356262\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "test_loss, test_acc = model.evaluate(test_images, test_labels)\n",
        "print('Test accuracy:', test_acc)\n",
        "\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.axhline(y=test_acc, color='r', linestyle='-')\n",
        "plt.title('Model Accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.legend(['Train', 'Validation', 'Test'], loc='upper left')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 228,
      "metadata": {
        "id": "DLLIrvy2psKZ"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# list all data in history\n",
        "print(history.history.keys())\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', 'validation'], loc='upper left')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 229,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# summarize history for loss\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()"
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {
        "id": "ImI7busBrCie"
      },
      "source": [
        "Additional questions:\n",
        "\n",
        "\n",
        "*   Describe whether you found any differences in the network’s accuracy when applying regularisation compared to not applying it. If there were differences, which regularisation did you use? If no differences were found, what could be the reason? **(10 marks)**\n",
        "\n",
        "*your answer here*\n",
        "\n",
        "The accuracy of the model was highest when using a combination of L1L2 regularisation. With no regularisation applied the accuracy was marginally lower than when using L2 or a combination of L1L2 and it acctually outperformed L1 regularization. This may suggest that the model did not have a significant problem with overfitting, which is what regularization is primarily used to address.\n",
        "\n",
        "The fact that L2 regularization did not perform significantly better than the no regularization case suggests that there may not have been a significant problem with the model's weights growing too large.\n",
        "\n",
        "**(Accuracy results)**\n",
        "\n",
        "l1l2 = 0.8828 acc 0.1137 loss\n",
        "\n",
        "no regularisation = 0.8773 acc 0.0599 loss\n",
        "\n",
        "L1 = 0.8761 acc 0.1057 loss\n",
        "\n",
        "L2 = 0.8778 acc 0.0924 loss \n"
      ]
    },
    {
      "attachments": {},
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Write your conclusions about the results achieved with your model on the fashion MNIST dataset and ideas to improve these results/performance further. \n",
        "\n",
        "**(10 marks)**\n",
        "\n",
        "\n",
        "Based on the results I obtained I can say that the model performs reasonably well with an accuracy of 0.8828 and loss of 0.1137. Suprisingly the application of regularisation did not make a huge impact and in fact the model had a lower loss value when no regularisation was applied. When regularization is applied, the model is penalized for having large weights, this can reduce the model's complexity and help prevent overfitting  leading to higher validation accuracy but at the cost of slightly higher loss.\n",
        "\n",
        "Something I find interesting is that the graphs above indicate that the model could be overfitting as the validation accuracy and loss are both higher than the training sets. However this is not what is shown by my model evaluation results. This could be because the model is not overfitting but rather the validation set is not representative of the training set. This could be because the validation set is too small and does not contain enough examples of each class. This could be improved by increasing the size of the validation set. \n",
        "\n",
        "One way of addressing this issue would be data augmentation which would increase the size of the training set by generating new images from the existing ones. This would help to prevent overfitting by increasing the size of the training set and also help to prevent underfitting by generating new images that are similar to the existing ones. This would also help to address the issue of the validation set not being representative of the training set as it would increase the size of the validation set.\n",
        "\n",
        "There is obvious room for improvment and more fine tuning of hyper parameters could help improve the models performance. I tried many different combinations of optimisers, regularisation and learning rates but I was unable to find the best solution. i believe that data augmentation is the key to breaking though this barrier and improving the models performance. Increasing epoch could improve accuracy but with diminishing returns as the model will eventually overfit.\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EcSc_1JiqRA3"
      },
      "source": [
        "Additional remarks:\n",
        "\n",
        "*   Code outline appropriately commented. **(10 marks)**\n",
        "*   Code running without errors. **(10 marks)**\n",
        "\n",
        "---\n",
        "\n"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "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": "3.10.4"
    },
    "vscode": {
      "interpreter": {
        "hash": "d57303a279247f119f3ff45d18d6802b0adc55d51f66ed848a267a8b72ebe61c"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}