03. Image Classification/StochasticGradientDescent.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "StochasticGradientDescent.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-GapUcfSfTCm"
      },
      "source": [
        "### **INITIALIZATION:**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IsqD9nTHe-Eu"
      },
      "source": [
        "#@ IMPORTING NECESSARY PACKAGES AND LIBRARIES: \n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import classification_report\n",
        "from sklearn.datasets import make_blobs\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1mib_bvkfRcS"
      },
      "source": [
        "#@ DEFINING SIGMOID ACTIVATION FUNCTION: \n",
        "def sigmoid_activation(x):\n",
        "    return 1.0 / (1 + np.exp(-x))\n",
        "\n",
        "#@ DEFINING PREDICT FUNCTION:\n",
        "def predict(X, W):                                     # Defining Function. \n",
        "    preds = sigmoid_activation(X.dot(W))               # Initializing Dot Product. \n",
        "    preds[preds <= 0.5] = 0                            # Implementing Thresholds. \n",
        "    preds[preds > 0.5] = 1                             # Implementing Thresholds. \n",
        "    return preds                                       # Getting Predictions. \n",
        "\n",
        "#@ DEFINING NEXT BATCH FUNCTION: \n",
        "def next_batch(X, y, batchSize):                       # Defining Function. \n",
        "    for i in np.arange(0, X.shape[0], batchSize):\n",
        "        yield (X[i:i + batchSize], y[i:i + batchSize]) # Initializing Batches. "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YILkUewBg9_P"
      },
      "source": [
        "**GETTING THE DATASET:**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yWzwqYOSfmWd"
      },
      "source": [
        "#@ GETTING THE DATASET: EXAMPLE: \n",
        "(X, y) = make_blobs(n_samples=1000, n_features=2, centers=2, \n",
        "                    cluster_std=1.5, random_state=42)           # Initializing Classification Dataset. \n",
        "y = y.reshape((y.shape[0], 1))                                  # Reshaping Dataset. \n",
        "X = np.c_[X, np.ones((X.shape[0]))]                             # Inserting Column Matrix. \n",
        "(trainX, testX, trainY, testY) = train_test_split(\n",
        "    X, y, test_size=0.5, random_state=42)                       # Partitioning Dataset into Training and Testing. "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bvHTV9bIwuyO"
      },
      "source": [
        "### **STOCHASTIC GRADIENT DESCENT:**\n",
        "- Stochastic Gradient Descent is a simple modification to the standard gradient descent algorithm that computes the gradient and updates the weight matrix on small batches of training data, rather than the entire training set. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b3al62kahFim",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "485ac757-1c9b-4736-bfa6-d016f57a9733"
      },
      "source": [
        "#@ INITIALIZING PARAMETERS: \n",
        "print(\"[INFO] training...\")\n",
        "epochs, lr, batchsize = 100, 0.01, 32                  # Initializing Epochs and LR. \n",
        "W = np.random.randn(X.shape[1], 1)                     # Initializing Weights. \n",
        "losses = []                                            # Initialization. \n",
        "\n",
        "#@ INITIALIZING STOCHASTIC GRADIENT DESCENT: \n",
        "for epoch in np.arange(0, epochs):\n",
        "    epochLoss = []                                     # Initialization. \n",
        "    for (batchX, batchY) in next_batch(X,y,batchsize):\n",
        "        preds = sigmoid_activation(batchX.dot(W))      # Implementating Sigmoid Activation. \n",
        "        error = preds - batchY                         # Computing Least Square Error. \n",
        "        epochLoss.append(np.sum(error**2))             # Updating. \n",
        "        gradient = batchX.T.dot(error)                 # Computing Gradient Descent. \n",
        "        W += -lr * gradient                            # Updating Weights. \n",
        "    loss = np.average(epochLoss)                       # Initializing Loss. \n",
        "    losses.append(loss)\n",
        "    if epoch==0 or (epoch+1) % 5 == 0:\n",
        "        print(\"[INFO] epoch={}, loss={:.7f}\".format(\n",
        "            int(epoch+1), loss))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO] training...\n",
            "[INFO] epoch=1, loss=1.2054598\n",
            "[INFO] epoch=5, loss=0.0792159\n",
            "[INFO] epoch=10, loss=0.0675258\n",
            "[INFO] epoch=15, loss=0.0616096\n",
            "[INFO] epoch=20, loss=0.0576062\n",
            "[INFO] epoch=25, loss=0.0545596\n",
            "[INFO] epoch=30, loss=0.0520902\n",
            "[INFO] epoch=35, loss=0.0500094\n",
            "[INFO] epoch=40, loss=0.0482092\n",
            "[INFO] epoch=45, loss=0.0466223\n",
            "[INFO] epoch=50, loss=0.0452035\n",
            "[INFO] epoch=55, loss=0.0439211\n",
            "[INFO] epoch=60, loss=0.0427516\n",
            "[INFO] epoch=65, loss=0.0416777\n",
            "[INFO] epoch=70, loss=0.0406856\n",
            "[INFO] epoch=75, loss=0.0397644\n",
            "[INFO] epoch=80, loss=0.0389054\n",
            "[INFO] epoch=85, loss=0.0381014\n",
            "[INFO] epoch=90, loss=0.0373465\n",
            "[INFO] epoch=95, loss=0.0366356\n",
            "[INFO] epoch=100, loss=0.0359644\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WJH4JfVm0Wjt",
        "outputId": "5057cdbe-7a67-44fe-c3c8-beb8ec5dad6a"
      },
      "source": [
        "#@ INITIALIZING MODEL EVALUATION: \n",
        "print(\"[INFO] evaluating...\")\n",
        "preds = predict(testX, W)                   # Computing Predictions. \n",
        "print(classification_report(testY, preds))  # Inspecting Classification Report. "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO] evaluating...\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       1.00      1.00      1.00       252\n",
            "           1       1.00      1.00      1.00       248\n",
            "\n",
            "    accuracy                           1.00       500\n",
            "   macro avg       1.00      1.00      1.00       500\n",
            "weighted avg       1.00      1.00      1.00       500\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "Z7NLu6tX01tF",
        "outputId": "54c5e2bc-b2a8-445f-e687-cbc81d1f8d27"
      },
      "source": [
        "#@ INSPECTING CLASSIFICATION DATA: \n",
        "plt.style.use(\"ggplot\")\n",
        "plt.figure()\n",
        "plt.title(\"Data\")\n",
        "plt.scatter(testX[:, 0], testX[:, 1], marker=\"o\", c=testY, s=30)\n",
        "plt.show();"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 299
        },
        "id": "Ls92-ZDy1WeC",
        "outputId": "ce96e95b-a474-4061-9199-7e139f9800ab"
      },
      "source": [
        "#@ INSPECTING LOSS:\n",
        "plt.style.use(\"ggplot\")\n",
        "plt.figure()\n",
        "plt.plot(np.arange(0, epochs), losses)\n",
        "plt.title(\"Training Loss\")\n",
        "plt.xlabel(\"Epoch\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.show();"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ScNm_RQAOFAv"
      },
      "source": [
        "### **IMAGE CLASSIFICATION**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "baBfMDKAO9h7"
      },
      "source": [
        "#@ INITIALIZING DEPENDENCIES: UNCOMMENT BELOW: \n",
        "# !python /content/simplepreprocessor.py"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NrqcxIdW16o9"
      },
      "source": [
        "#@ IMPORTING NECESSARY PACKAGES AND DEPENDENCIES: \n",
        "from sklearn.linear_model import SGDClassifier\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "from sklearn.model_selection import train_test_split\n",
        "from simplepreprocessor import SimplePreprocessor\n",
        "from simplepreprocessor import SimpleDatasetLoader\n",
        "from imutils import paths"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2RbkOkwcPeJR"
      },
      "source": [
        "#@ GETTING DATASET: UNCOMMENT BELOW: \n",
        "# !unzip /content/drive/MyDrive/Data/animals.zip"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-1yEpufuP7u3",
        "outputId": "fb7b0fa5-a753-4445-cdaf-b9696b4851e3"
      },
      "source": [
        "#@ PREPROCESSING DATASET: \n",
        "imagePaths = list(paths.list_images(\"/content/animals\"))         # Getting Image Paths. \n",
        "sp = SimplePreprocessor(32, 32)                                  # Initializing Image Resizing. \n",
        "sdl = SimpleDatasetLoader(preprocessors=[sp])                    # Initializing Dataset Loader. \n",
        "(data, labels) = sdl.load(imagePaths, verbose=500)               # Initializing Data and Labels. \n",
        "data = data.reshape((data.shape[0], 3072))                       # Initializing Flatten. \n",
        "\n",
        "#@ ENCODING LABELS: \n",
        "le = LabelEncoder()                                              # Initializing Label Encoder. \n",
        "labels = le.fit_transform(labels)                                # Encoding Labels as Integers. \n",
        "(trainX, testX, trainY, testY) = train_test_split(\n",
        "    data, labels, test_size=0.25, random_state=42)               # Initializing Training and Testing. \n",
        "\n",
        "#@ IMPLEMENTING SGDCLASSIFIER AND REGULARIZATION: \n",
        "for r in (None, \"l1\", \"l2\"):\n",
        "    print(\"[INFO] training model with {} penalty\".format(r))\n",
        "    model = SGDClassifier(loss=\"log\", penalty=r, max_iter=100, \n",
        "                          learning_rate=\"constant\", eta0=0.01, \n",
        "                          random_state=42)                       # Initializing SGD Classier. \n",
        "    model.fit(trainX, trainY)                                    # Training SGD Classifier. \n",
        "    acc = model.score(testX, testY)                              # Getting Accuracy. \n",
        "    print(\"[INFO] {} penalty accuracy: {:.2f}%\".format(\n",
        "        r, acc*100))"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO] processed 500/3000\n",
            "[INFO] processed 1000/3000\n",
            "[INFO] processed 1500/3000\n",
            "[INFO] processed 2000/3000\n",
            "[INFO] processed 2500/3000\n",
            "[INFO] processed 3000/3000\n",
            "[INFO] training model with None penalty\n",
            "[INFO] None penalty accuracy: 49.07%\n",
            "[INFO] training model with l1 penalty\n",
            "[INFO] l1 penalty accuracy: 50.80%\n",
            "[INFO] training model with l2 penalty\n",
            "[INFO] l2 penalty accuracy: 49.47%\n"
          ]
        }
      ]
    }
  ]
}