XOR_MLP_problem.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "XOR MLP problem.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "Shibu Mohapatra\n",
        "\n",
        "MSC CS Artificial Intelligence"
      ],
      "metadata": {
        "id": "IDGVUX9hlWk6"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IMH2FkL6EH0S"
      },
      "source": [
        "**Program to solve XOR Problem. Solution to XOR in (MLP) Multi-Layer Perceptron.**"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "n_x = 2 # number of neurons in input layer\n",
        "n_h = 8 # number of neurons in hidden layer\n",
        "n_y = 1 # number of neurons in output layer\n",
        "\n",
        "X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).T\n",
        "y = np.array([0, 1, 1, 0])\n",
        "print(X.shape)\n",
        "print(y.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uZv4RrceW3zs",
        "outputId": "91d0082c-4cee-4391-9589-1e0e9171a8a5"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(2, 4)\n",
            "(4,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(X)\n",
        "print(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rwAbJQmkXBop",
        "outputId": "62f8417b-687e-4a65-902d-9e8c4c1f7289"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[0 0 1 1]\n",
            " [0 1 0 1]]\n",
            "[0 1 1 0]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def forward_prop(X, y, params):\n",
        "  W1 = params['W1']\n",
        "  b1 = params['b1']\n",
        "  W2 = params['W2']\n",
        "  b2 = params['b2']\n",
        "\n",
        "  m = X.shape[0]\n",
        "\n",
        "  Z1 = np.dot(W1, X) + b1\n",
        "  A1 = np.maximum(0, Z1)\n",
        "\n",
        "  Z2 = np.dot(W2, A1) + b2\n",
        "  A2 = 1/(1 + np.exp(-Z2))\n",
        "\n",
        "  loss = (-1 / m) * np.sum(y * np.log(A2) + (1-y) * np.log(1-A2))\n",
        "\n",
        "  return Z1, A1, Z2, A2, loss"
      ],
      "metadata": {
        "id": "vqwaqt4kXqkG"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def backward_prop(X, y, params):\n",
        "\n",
        "  W1 = params['W1']\n",
        "  b1 = params['b1']\n",
        "  W2 = params['W2']\n",
        "  b2 = params['b2']\n",
        "\n",
        "  Z1, A1, Z2, A2, loss = forward_prop(X, y, params)\n",
        "\n",
        "  dloss_A2 = - (np.divide(y, A2) - np.divide(1 - y, 1 - A2)) \n",
        "  dloss_Z2 = dloss_A2 * (A2 * (1 - A2))\n",
        "  dloss_A1 = np.dot(W2.T, dloss_Z2)\n",
        "  dloss_W2 = (1/A1.shape[1]) * np.dot(dloss_Z2, A1.T)\n",
        "  dloss_b2 = (1/A1.shape[1]) * np.dot(dloss_Z2, np.ones([dloss_Z2.shape[1],1]))\n",
        "\n",
        "  dloss_Z1 = dloss_A1 * ((Z1 >= 0) * 1)\n",
        "  dloss_A0 = np.dot(W1.T, dloss_Z1)\n",
        "  dloss_W1 = (1/X.shape[1]) * np.dot(dloss_Z1, X.T)\n",
        "  dloss_b1 = (1/X.shape[1]) * np.dot(dloss_Z1, np.ones([dloss_Z1.shape[1],1]))\n",
        "\n",
        "  grads = {}\n",
        "  grads['dloss_W1'] = dloss_W1\n",
        "  grads['dloss_b1'] = dloss_b1\n",
        "  grads['dloss_W2'] = dloss_W2\n",
        "  grads['dloss_b2'] = dloss_b2\n",
        "\n",
        "  return grads, loss"
      ],
      "metadata": {
        "id": "TjVG7l9iXtl7"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def train_nn(X, y, learning_rate = 0.01, epochs = 250):\n",
        "\n",
        "  np.random.seed(1)\n",
        "  W1 = np.random.rand(n_h, n_x)\n",
        "  b1 = np.zeros((n_h, 1))\n",
        "\n",
        "  W2 = np.random.randn(n_y, n_h)\n",
        "  b2 = np.zeros((n_y, 1))\n",
        "  params = {'W1': W1, 'b1': b1, 'W2': W2, 'b2': b2}\n",
        "\n",
        "  losses = []\n",
        "  lst_w1 = []\n",
        "  for epoch in range(epochs):\n",
        "    grads, loss = backward_prop(X, y, params)\n",
        "    losses.append(loss)\n",
        "\n",
        "    params['W1'] = params['W1'] - learning_rate * grads['dloss_W1']\n",
        "    params['b1'] = params['b1'] - learning_rate * grads['dloss_b1']\n",
        "    params['W2'] = params['W2'] - learning_rate * grads['dloss_W2']\n",
        "    params['b2'] = params['b2'] - learning_rate * grads['dloss_b2']\n",
        "\n",
        "  return params, losses"
      ],
      "metadata": {
        "id": "1vXgZ4BKXwAJ"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "params, losses = train_nn(X, y, learning_rate = 0.1, epochs = 5000)\n",
        "fig, ax = plt.subplots(figsize = (12, 4))\n",
        "ax.plot(losses)\n",
        "ax.set_title('Training Loss')\n",
        "ax.set_xlabel('Number of Epochs')\n",
        "ax.set_ylabel('Cross Entropy Loss')\n",
        "plt.show()\n",
        "_, _, _, preds, loss = forward_prop(X, y, params)\n",
        "print('Ground Truth:', y)\n",
        "print('Predictions:', (preds.reshape(-1) >= 0.5) * 1)\n",
        "print('Prediction Probabilities:', preds.reshape(-1))\n",
        "print('Loss:', loss)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 365
        },
        "id": "Ltz28nEzXyQD",
        "outputId": "d3907cf8-0f4d-444a-b66f-ff161821f7aa"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Ground Truth: [0 1 1 0]\n",
            "Predictions: [0 1 1 0]\n",
            "Prediction Probabilities: [0.01184645 0.9978798  0.9978594  0.00148388]\n",
            "Loss: 0.008833749991272684\n"
          ]
        }
      ]
    }
  ]
}