gcn.ipynb

gcn.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyMBVgZrUrJethoUF/5Gf2Id",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/AhmedCoolProjects/6344dd2d104fc068d05cf50b5b08f828/gcn.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "9ru-uTBrygoy"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "\n",
        "class GCNLayer(nn.Module):\n",
        "    def __init__(self, in_feats, out_feats, use_relu=True):\n",
        "        super().__init__()\n",
        "        # Define the learnable weights here\n",
        "        self.linear = nn.Linear(in_feats, out_feats) # this will create W and b\n",
        "        self.use_relu = use_relu\n",
        "\n",
        "    def forward(self, node_feats, adj_matrix):\n",
        "        # 1. transform\n",
        "        trans_feats = self.linear(node_feats)\n",
        "        # 2. normalization\n",
        "        I = torch.eye(adj_matrix.shape[0])\n",
        "        A_tilde = adj_matrix + I\n",
        "        D_tilde_sqrt = torch.diag(torch.pow(A_tilde.sum(dim=1), -0.5))\n",
        "        A_norm = D_tilde_sqrt @ A_tilde @ D_tilde_sqrt\n",
        "        # 3. message passing\n",
        "        message = A_norm @ trans_feats\n",
        "        # 4. update\n",
        "        if self.use_relu:\n",
        "            return F.relu(message)\n",
        "        return message\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class GCN(nn.Module):\n",
        "    def __init__(self, input_dim, hidden_dim, output_dim):\n",
        "        super().__init__()\n",
        "        # Layer 1\n",
        "        self.gcn1 = GCNLayer(input_dim, hidden_dim, use_relu=True)\n",
        "        # Layer 2\n",
        "        self.gcn2 = GCNLayer(hidden_dim, output_dim, use_relu=False)\n",
        "\n",
        "    def forward(self, node_feats, adj_matrix):\n",
        "        x = self.gcn1(node_feats, adj_matrix)\n",
        "        x = self.gcn2(x, adj_matrix)\n",
        "        return x\n"
      ],
      "metadata": {
        "id": "ahFWzCGAymtS"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# --- Dummy Data ---\n",
        "# 4 Nodes, 3 Features each\n",
        "node_features = torch.tensor([\n",
        "    [1.0, 0.0, 0.0], # Node 0\n",
        "    [0.0, 1.0, 0.0], # Node 1\n",
        "    [0.0, 0.0, 1.0], # Node 2\n",
        "    [1.0, 1.0, 0.0]  # Node 3\n",
        "])\n",
        "\n",
        "# Adjacency Matrix (4x4)\n",
        "adj_matrix = torch.tensor([\n",
        "    [0, 1, 1, 0],\n",
        "    [1, 0, 1, 0],\n",
        "    [1, 1, 0, 1],\n",
        "    [0, 0, 1, 0]\n",
        "], dtype=torch.float32)\n",
        "\n",
        "# Labels: We want to classify the nodes into Class 0 or Class 1\n",
        "# Let's say Node 0 & 1 are Class 0, Node 2 & 3 are Class 1\n",
        "labels = torch.tensor([0, 0, 1, 1])\n",
        "\n",
        "# --- Initialize Model ---\n",
        "# Input: 3 features -> Hidden: 4 features -> Output: 2 classes\n",
        "model = GCN(input_dim=3, hidden_dim=4, output_dim=2)\n",
        "\n",
        "# --- Optimizer & Loss ---\n",
        "# Optimizer handles the parameter updates (Gradient Descent)\n",
        "optimizer = optim.Adam(model.parameters(), lr=0.01)\n",
        "# CrossEntropyLoss is standard for classification\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "\n",
        "print(\"Starting Training...\\n\")\n",
        "\n",
        "for epoch in range(100): # Run for 100 loops\n",
        "    model.train() # Set model to training mode\n",
        "\n",
        "    # 1. Zero Gradients\n",
        "    # Clear old gradients from the previous step\n",
        "    optimizer.zero_grad()\n",
        "\n",
        "    # 2. Forward Pass\n",
        "    # Get predictions from the model\n",
        "    output = model(node_features, adj_matrix)\n",
        "\n",
        "    # 3. Calculate Loss\n",
        "    # Compare output with actual labels\n",
        "    loss = criterion(output, labels)\n",
        "\n",
        "    # 4. Backward Pass (The Math Magic)\n",
        "    # PyTorch calculates gradients (dLoss/dW) for all weights automatically\n",
        "    loss.backward()\n",
        "\n",
        "    # 5. Update Weights\n",
        "    # Adjust weights: W_new = W_old - (lr * gradient)\n",
        "    optimizer.step()\n",
        "\n",
        "    if epoch % 10 == 0:\n",
        "        print(f\"Epoch {epoch} | Loss: {loss.item():.4f}\")\n",
        "\n",
        "# Final Prediction\n",
        "print(\"\\nFinal Node Classifications:\")\n",
        "final_output = model(node_features, adj_matrix)\n",
        "predicted_classes = final_output.argmax(dim=1)\n",
        "print(f\"Predicted: {predicted_classes}\")\n",
        "print(f\"Actual:    {labels}\")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LmrcAT_7ypw4",
        "outputId": "bbd9408b-b928-4b32-94c0-1c1c07243d8f"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Starting Training...\n",
            "\n",
            "Epoch 0 | Loss: 0.6901\n",
            "Epoch 10 | Loss: 0.6806\n",
            "Epoch 20 | Loss: 0.6728\n",
            "Epoch 30 | Loss: 0.6637\n",
            "Epoch 40 | Loss: 0.6526\n",
            "Epoch 50 | Loss: 0.6377\n",
            "Epoch 60 | Loss: 0.6215\n",
            "Epoch 70 | Loss: 0.6020\n",
            "Epoch 80 | Loss: 0.5817\n",
            "Epoch 90 | Loss: 0.5586\n",
            "\n",
            "Final Node Classifications:\n",
            "Predicted: tensor([0, 0, 1, 1])\n",
            "Actual:    tensor([0, 0, 1, 1])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "GKn-lXi6ysL4"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}