485_Week_8.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "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/github/gtsagkatakis/Data-Science-and-Applications/blob/main/485_Week_8.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Introduction to Graph Neural Networks"
      ],
      "metadata": {
        "id": "SbbW-ULMzaff"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Graph representation\n",
        "\n",
        "Mathematically, a graph {G} is defined as a tuple of a set of nodes/vertices V, and a set of edges/links E: {G}=(V,E). Each edge is a pair of two vertices, and represents a connection between them. For instance, let’s look at the following graph:\n",
        "\n",
        "![](https://github.com/Lightning-AI/lightning-tutorials/raw/main/course_UvA-DL/06-graph-neural-networks/example_graph.svg)\n",
        "\n",
        "\n",
        "The vertices are V=\\{1,2,3,4\\}, and edges E=\\{(1,2), (2,3), (2,4), (3,4)\\}. Note that for simplicity, we assume the graph to be undirected and hence don’t add mirrored pairs like (2,1). In application, vertices and edge can often have specific attributes, and edges can even be directed.\n",
        "\n",
        "The question is how we could represent this diversity in an efficient way for matrix operations. Usually, for the edges, we decide between two variants: an adjacency matrix, or a list of paired vertex indices.\n",
        "\n",
        "The adjacency matrix A is a square matrix whose elements indicate whether pairs of vertices are adjacent, i.e. connected, or not. In the simplest case, $A_{ij}$ is 1 if there is a connection from node i to j, and otherwise 0. If we have edge attributes or different categories of edges in a graph, this information can be added to the matrix as well. For an undirected graph, keep in mind that A is a symmetric matrix $(A_{ij}=A_{ji})$. For the example graph above, we have the following adjacency matrix:\n",
        "\n",
        "$$A = \\begin{bmatrix}\n",
        "    0 & 1 & 0 & 0\\\\\n",
        "    1 & 0 & 1 & 1\\\\\n",
        "    0 & 1 & 0 & 1\\\\\n",
        "    0 & 1 & 1 & 0\n",
        "\\end{bmatrix}$$\n",
        "\n",
        "While expressing a graph as a list of edges is more efficient in terms of memory and (possibly) computation, using an adjacency matrix is more intuitive and simpler to implement. Alternatively, we could also use the list of edges to define a sparse adjacency matrix with which we can work as if it was a dense matrix, but allows more memory-efficient operations.\n"
      ],
      "metadata": {
        "id": "ByU_DiD7zi2-"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Graph Convolutional Networks\n",
        "\n",
        "Graph Convolutional Networks (GCNs) are similar to convolutions in images in the sense that the “filter” parameters are typically shared over all locations in the graph. At the same time, GCNs rely on message passing methods, which means that vertices exchange information with the neighbors, and send “messages” to each other.\n",
        "\n",
        "Let's try to visually understand how GCNs work.\n",
        "- The first step is that each node creates a feature vector that represents the message it wants to send to all its neighbors.\n",
        "- In the second step, the messages are sent to the neighbors, so that a node receives one message per adjacent node.\n",
        "- Finally, the messages generated by neighboring nodes are aggregated to create a new representation for each node. This aggregation operation typically involves summing, averaging, or concatenating the messages, often with learnable parameters.\n",
        "\n",
        "Below we have visualized the two steps for our example graph.\n",
        "\n",
        "![](https://github.com/Lightning-AI/lightning-tutorials/raw/main/course_UvA-DL/06-graph-neural-networks/graph_message_passing.svg)"
      ],
      "metadata": {
        "id": "0Lwyw2k21ntM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Implementation in Pytorch\n",
        "\n",
        "Ιmplementing graph networks with adjacency matrix is simple and straight-forward but can be computationally expensive for large graphs. Many real-world graphs can reach over 200K nodes, for which adjacency matrix-based implementations fail. There are a lot of optimizations possible when implementing GNNs, and luckily, there exist packages that provide such layers.\n",
        "\n",
        "PyTorch Geometric simplifies deep learning on graph data by providing efficient tools and ready-to-use modules for tasks like:\n",
        "- graph/node classification,\n",
        "- link prediction,\n",
        "- graph generation.\n",
        "\n",
        "It also offers benchmark datasets and includes various preprocessing tools for common graph-related tasks such as data augmentation, feature engineering, and graph transformation."
      ],
      "metadata": {
        "id": "NF56caXz2HId"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Install required packages.\n",
        "import os\n",
        "import torch\n",
        "os.environ['TORCH'] = torch.__version__\n",
        "print(torch.__version__)\n",
        "\n",
        "!pip install -q torch-scatter -f https://data.pyg.org/whl/torch-${TORCH}.html\n",
        "!pip install -q torch-sparse -f https://data.pyg.org/whl/torch-${TORCH}.html\n",
        "!pip install -q git+https://github.com/pyg-team/pytorch_geometric.git"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FyRBXAq4-Hzb",
        "outputId": "9708b76d-8f4d-4603-d3ac-48672e3124a2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2.2.1+cu121\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m10.9/10.9 MB\u001b[0m \u001b[31m25.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m5.0/5.0 MB\u001b[0m \u001b[31m28.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25h  Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n",
            "  Getting requirements to build wheel ... \u001b[?25l\u001b[?25hdone\n",
            "  Preparing metadata (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n",
            "  Building wheel for torch_geometric (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## A Practical Example: Node Classification with Graph Neural Networks\n",
        "\n",
        "\n",
        "Node-level tasks have the goal to classify nodes in a graph. Usually, we have given a single, large graph with more than 1000 nodes of which a certain amount of nodes are labeled. We learn to classify those labeled examples during training and try to generalize to the unlabeled nodes.\n",
        "\n",
        "\n",
        "![](https://www.researchgate.net/publication/355873169/figure/fig1/AS:1119949498126336@1644028176568/Illustration-of-semi-supervised-node-classification-Blue-and-green-denote-those-nodes_W640.jpg)"
      ],
      "metadata": {
        "id": "XzrWWA2zCSOw"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Load the Dataset\n",
        "\n",
        "A popular example that we will use in this tutorial is the Cora dataset, a citation network among papers. The Cora consists of 2708 scientific publications with links between each other representing the citation of one paper by another. The task is to classify each publication into one of seven classes."
      ],
      "metadata": {
        "id": "81ORpoZ580hU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from torch_geometric.datasets import Planetoid\n",
        "from torch_geometric.transforms import NormalizeFeatures\n",
        "\n",
        "dataset = Planetoid(root='data/Planetoid', name='Cora', transform=NormalizeFeatures())\n",
        "\n",
        "print()\n",
        "print(f'Dataset: {dataset}:')\n",
        "print('======================')\n",
        "print(f'Number of graphs: {len(dataset)}')\n",
        "print(f'Number of features: {dataset.num_features}')\n",
        "print(f'Number of classes: {dataset.num_classes}')\n",
        "\n",
        "data = dataset[0]  # Get the first graph object.\n",
        "\n",
        "print()\n",
        "print(data)\n",
        "print('===========================================================================================================')\n",
        "\n",
        "# Gather some statistics about the graph.\n",
        "print(f'Number of nodes: {data.num_nodes}')\n",
        "print(f'Number of edges: {data.num_edges}')\n",
        "print(f'Average node degree: {data.num_edges / data.num_nodes:.2f}')\n",
        "print(f'Number of training nodes: {data.train_mask.sum()}')\n",
        "print(f'Training node label rate: {int(data.train_mask.sum()) / data.num_nodes:.2f}')\n",
        "print(f'Has isolated nodes: {data.has_isolated_nodes()}')\n",
        "print(f'Has self-loops: {data.has_self_loops()}')\n",
        "print(f'Is undirected: {data.is_undirected()}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IA0ZJ1yo0I29",
        "outputId": "491b0557-bba8-420a-d556-ca98310c674c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.x\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.tx\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.allx\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.y\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.ty\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.ally\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.graph\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.test.index\n",
            "Processing...\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Dataset: Cora():\n",
            "======================\n",
            "Number of graphs: 1\n",
            "Number of features: 1433\n",
            "Number of classes: 7\n",
            "\n",
            "Data(x=[2708, 1433], edge_index=[2, 10556], y=[2708], train_mask=[2708], val_mask=[2708], test_mask=[2708])\n",
            "===========================================================================================================\n",
            "Number of nodes: 2708\n",
            "Number of edges: 10556\n",
            "Average node degree: 3.90\n",
            "Number of training nodes: 140\n",
            "Training node label rate: 0.05\n",
            "Has isolated nodes: False\n",
            "Has self-loops: False\n",
            "Is undirected: True\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Done!\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "![](https://miro.medium.com/v2/resize:fit:1100/format:webp/1*oygeCjtUsS87duvFoDT8tA.png)"
      ],
      "metadata": {
        "id": "OelEO7fEEkKI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Let's first train a Multi-layer Perception Network (MLP)\n",
        "\n",
        "We can infer the category of a document solely based on its feature representation using a simple MLP. The MLP model operates on input node features using shared weights across all nodes."
      ],
      "metadata": {
        "id": "sHp7rabGWkVP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "from torch.nn import Linear\n",
        "import torch.nn.functional as F\n",
        "\n",
        "\n",
        "class MLP(torch.nn.Module):\n",
        "    def __init__(self, hidden_channels):\n",
        "        super().__init__()\n",
        "        torch.manual_seed(12345)\n",
        "        self.lin1 = Linear(dataset.num_features, hidden_channels)\n",
        "        self.lin2 = Linear(hidden_channels, dataset.num_classes)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.lin1(x)\n",
        "        x = x.relu()\n",
        "        x = F.dropout(x, p=0.5, training=self.training)\n",
        "        x = self.lin2(x)\n",
        "        return x\n",
        "\n",
        "model = MLP(hidden_channels=16)\n",
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cz4MZjBTWjzC",
        "outputId": "9a3898aa-3551-4211-b327-fadb7068b1ee"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "MLP(\n",
            "  (lin1): Linear(in_features=1433, out_features=16, bias=True)\n",
            "  (lin2): Linear(in_features=16, out_features=7, bias=True)\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Our MLP is defined by two linear layers and enhanced by ReLU non-linearity and dropout. Here, we first reduce the 1433-dimensional feature vector to a low-dimensional embedding (hidden_channels=16), while the second linear layer acts as a classifier that should map each low-dimensional node embedding to one of the 7 classes.\n",
        "\n",
        "Let's train our simple MLP. We use the cross entropy loss and Adam optimizer. This time, we also define a test function to evaluate how well our final model performs on the test node set (which labels have not been observed during training)."
      ],
      "metadata": {
        "id": "xbMwKQreWso2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model = MLP(hidden_channels=16)\n",
        "criterion = torch.nn.CrossEntropyLoss()  # Define loss criterion.\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=0.01, weight_decay=5e-4)  # Define optimizer.\n",
        "\n",
        "def train():\n",
        "      model.train()\n",
        "      optimizer.zero_grad()  # Clear gradients.\n",
        "      out = model(data.x)  # Perform a single forward pass.\n",
        "      loss = criterion(out[data.train_mask], data.y[data.train_mask])  # Compute the loss solely based on the training nodes.\n",
        "      loss.backward()  # Derive gradients.\n",
        "      optimizer.step()  # Update parameters based on gradients.\n",
        "      return loss\n",
        "\n",
        "def test():\n",
        "      model.eval()\n",
        "      out = model(data.x)\n",
        "      pred = out.argmax(dim=1)  # Use the class with highest probability.\n",
        "      test_correct = pred[data.test_mask] == data.y[data.test_mask]  # Check against ground-truth labels.\n",
        "      test_acc = int(test_correct.sum()) / int(data.test_mask.sum())  # Derive ratio of correct predictions.\n",
        "      return test_acc\n",
        "\n",
        "for epoch in range(1, 201):\n",
        "    loss = train()\n",
        "    print(f'Epoch: {epoch:03d}, Loss: {loss:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ORe0iB4UWsOx",
        "outputId": "e4e8b6db-3f13-4079-8ee6-31736b3fdf5c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch: 001, Loss: 1.9615\n",
            "Epoch: 002, Loss: 1.9557\n",
            "Epoch: 003, Loss: 1.9505\n",
            "Epoch: 004, Loss: 1.9423\n",
            "Epoch: 005, Loss: 1.9327\n",
            "Epoch: 006, Loss: 1.9279\n",
            "Epoch: 007, Loss: 1.9144\n",
            "Epoch: 008, Loss: 1.9087\n",
            "Epoch: 009, Loss: 1.9023\n",
            "Epoch: 010, Loss: 1.8893\n",
            "Epoch: 011, Loss: 1.8776\n",
            "Epoch: 012, Loss: 1.8594\n",
            "Epoch: 013, Loss: 1.8457\n",
            "Epoch: 014, Loss: 1.8365\n",
            "Epoch: 015, Loss: 1.8280\n",
            "Epoch: 016, Loss: 1.7965\n",
            "Epoch: 017, Loss: 1.7984\n",
            "Epoch: 018, Loss: 1.7832\n",
            "Epoch: 019, Loss: 1.7495\n",
            "Epoch: 020, Loss: 1.7441\n",
            "Epoch: 021, Loss: 1.7188\n",
            "Epoch: 022, Loss: 1.7124\n",
            "Epoch: 023, Loss: 1.6785\n",
            "Epoch: 024, Loss: 1.6660\n",
            "Epoch: 025, Loss: 1.6119\n",
            "Epoch: 026, Loss: 1.6236\n",
            "Epoch: 027, Loss: 1.5827\n",
            "Epoch: 028, Loss: 1.5784\n",
            "Epoch: 029, Loss: 1.5524\n",
            "Epoch: 030, Loss: 1.5020\n",
            "Epoch: 031, Loss: 1.5065\n",
            "Epoch: 032, Loss: 1.4742\n",
            "Epoch: 033, Loss: 1.4581\n",
            "Epoch: 034, Loss: 1.4246\n",
            "Epoch: 035, Loss: 1.4131\n",
            "Epoch: 036, Loss: 1.4112\n",
            "Epoch: 037, Loss: 1.3923\n",
            "Epoch: 038, Loss: 1.3055\n",
            "Epoch: 039, Loss: 1.2982\n",
            "Epoch: 040, Loss: 1.2543\n",
            "Epoch: 041, Loss: 1.2244\n",
            "Epoch: 042, Loss: 1.2331\n",
            "Epoch: 043, Loss: 1.1984\n",
            "Epoch: 044, Loss: 1.1796\n",
            "Epoch: 045, Loss: 1.1093\n",
            "Epoch: 046, Loss: 1.1284\n",
            "Epoch: 047, Loss: 1.1229\n",
            "Epoch: 048, Loss: 1.0383\n",
            "Epoch: 049, Loss: 1.0439\n",
            "Epoch: 050, Loss: 1.0563\n",
            "Epoch: 051, Loss: 0.9893\n",
            "Epoch: 052, Loss: 1.0508\n",
            "Epoch: 053, Loss: 0.9343\n",
            "Epoch: 054, Loss: 0.9639\n",
            "Epoch: 055, Loss: 0.8929\n",
            "Epoch: 056, Loss: 0.8705\n",
            "Epoch: 057, Loss: 0.9176\n",
            "Epoch: 058, Loss: 0.9239\n",
            "Epoch: 059, Loss: 0.8641\n",
            "Epoch: 060, Loss: 0.8578\n",
            "Epoch: 061, Loss: 0.7908\n",
            "Epoch: 062, Loss: 0.7856\n",
            "Epoch: 063, Loss: 0.7683\n",
            "Epoch: 064, Loss: 0.7816\n",
            "Epoch: 065, Loss: 0.7356\n",
            "Epoch: 066, Loss: 0.6951\n",
            "Epoch: 067, Loss: 0.7300\n",
            "Epoch: 068, Loss: 0.6939\n",
            "Epoch: 069, Loss: 0.7550\n",
            "Epoch: 070, Loss: 0.6864\n",
            "Epoch: 071, Loss: 0.7094\n",
            "Epoch: 072, Loss: 0.7238\n",
            "Epoch: 073, Loss: 0.7150\n",
            "Epoch: 074, Loss: 0.6191\n",
            "Epoch: 075, Loss: 0.6770\n",
            "Epoch: 076, Loss: 0.6487\n",
            "Epoch: 077, Loss: 0.6258\n",
            "Epoch: 078, Loss: 0.5821\n",
            "Epoch: 079, Loss: 0.5637\n",
            "Epoch: 080, Loss: 0.6368\n",
            "Epoch: 081, Loss: 0.6333\n",
            "Epoch: 082, Loss: 0.6434\n",
            "Epoch: 083, Loss: 0.5974\n",
            "Epoch: 084, Loss: 0.6176\n",
            "Epoch: 085, Loss: 0.5972\n",
            "Epoch: 086, Loss: 0.4690\n",
            "Epoch: 087, Loss: 0.6362\n",
            "Epoch: 088, Loss: 0.6118\n",
            "Epoch: 089, Loss: 0.5248\n",
            "Epoch: 090, Loss: 0.5520\n",
            "Epoch: 091, Loss: 0.6130\n",
            "Epoch: 092, Loss: 0.5361\n",
            "Epoch: 093, Loss: 0.5594\n",
            "Epoch: 094, Loss: 0.5049\n",
            "Epoch: 095, Loss: 0.5043\n",
            "Epoch: 096, Loss: 0.5235\n",
            "Epoch: 097, Loss: 0.5451\n",
            "Epoch: 098, Loss: 0.5329\n",
            "Epoch: 099, Loss: 0.5008\n",
            "Epoch: 100, Loss: 0.5350\n",
            "Epoch: 101, Loss: 0.5343\n",
            "Epoch: 102, Loss: 0.5138\n",
            "Epoch: 103, Loss: 0.5377\n",
            "Epoch: 104, Loss: 0.5353\n",
            "Epoch: 105, Loss: 0.5176\n",
            "Epoch: 106, Loss: 0.5229\n",
            "Epoch: 107, Loss: 0.4558\n",
            "Epoch: 108, Loss: 0.4883\n",
            "Epoch: 109, Loss: 0.4659\n",
            "Epoch: 110, Loss: 0.4908\n",
            "Epoch: 111, Loss: 0.4966\n",
            "Epoch: 112, Loss: 0.4725\n",
            "Epoch: 113, Loss: 0.4787\n",
            "Epoch: 114, Loss: 0.4390\n",
            "Epoch: 115, Loss: 0.4199\n",
            "Epoch: 116, Loss: 0.4810\n",
            "Epoch: 117, Loss: 0.4484\n",
            "Epoch: 118, Loss: 0.5080\n",
            "Epoch: 119, Loss: 0.4241\n",
            "Epoch: 120, Loss: 0.4745\n",
            "Epoch: 121, Loss: 0.4651\n",
            "Epoch: 122, Loss: 0.4652\n",
            "Epoch: 123, Loss: 0.5580\n",
            "Epoch: 124, Loss: 0.4861\n",
            "Epoch: 125, Loss: 0.4405\n",
            "Epoch: 126, Loss: 0.4292\n",
            "Epoch: 127, Loss: 0.4409\n",
            "Epoch: 128, Loss: 0.3575\n",
            "Epoch: 129, Loss: 0.4468\n",
            "Epoch: 130, Loss: 0.4603\n",
            "Epoch: 131, Loss: 0.4108\n",
            "Epoch: 132, Loss: 0.4601\n",
            "Epoch: 133, Loss: 0.4258\n",
            "Epoch: 134, Loss: 0.3852\n",
            "Epoch: 135, Loss: 0.4028\n",
            "Epoch: 136, Loss: 0.4245\n",
            "Epoch: 137, Loss: 0.4300\n",
            "Epoch: 138, Loss: 0.4693\n",
            "Epoch: 139, Loss: 0.4314\n",
            "Epoch: 140, Loss: 0.4031\n",
            "Epoch: 141, Loss: 0.4290\n",
            "Epoch: 142, Loss: 0.4110\n",
            "Epoch: 143, Loss: 0.3863\n",
            "Epoch: 144, Loss: 0.4215\n",
            "Epoch: 145, Loss: 0.4519\n",
            "Epoch: 146, Loss: 0.3940\n",
            "Epoch: 147, Loss: 0.4429\n",
            "Epoch: 148, Loss: 0.3527\n",
            "Epoch: 149, Loss: 0.4390\n",
            "Epoch: 150, Loss: 0.4212\n",
            "Epoch: 151, Loss: 0.4128\n",
            "Epoch: 152, Loss: 0.3779\n",
            "Epoch: 153, Loss: 0.4801\n",
            "Epoch: 154, Loss: 0.4130\n",
            "Epoch: 155, Loss: 0.3962\n",
            "Epoch: 156, Loss: 0.4262\n",
            "Epoch: 157, Loss: 0.4210\n",
            "Epoch: 158, Loss: 0.4081\n",
            "Epoch: 159, Loss: 0.4066\n",
            "Epoch: 160, Loss: 0.3782\n",
            "Epoch: 161, Loss: 0.3836\n",
            "Epoch: 162, Loss: 0.4172\n",
            "Epoch: 163, Loss: 0.3993\n",
            "Epoch: 164, Loss: 0.4477\n",
            "Epoch: 165, Loss: 0.3714\n",
            "Epoch: 166, Loss: 0.3610\n",
            "Epoch: 167, Loss: 0.4546\n",
            "Epoch: 168, Loss: 0.4387\n",
            "Epoch: 169, Loss: 0.3793\n",
            "Epoch: 170, Loss: 0.3704\n",
            "Epoch: 171, Loss: 0.4286\n",
            "Epoch: 172, Loss: 0.4131\n",
            "Epoch: 173, Loss: 0.3795\n",
            "Epoch: 174, Loss: 0.4230\n",
            "Epoch: 175, Loss: 0.4139\n",
            "Epoch: 176, Loss: 0.3586\n",
            "Epoch: 177, Loss: 0.3588\n",
            "Epoch: 178, Loss: 0.3911\n",
            "Epoch: 179, Loss: 0.3810\n",
            "Epoch: 180, Loss: 0.4203\n",
            "Epoch: 181, Loss: 0.3583\n",
            "Epoch: 182, Loss: 0.3690\n",
            "Epoch: 183, Loss: 0.4025\n",
            "Epoch: 184, Loss: 0.3920\n",
            "Epoch: 185, Loss: 0.4369\n",
            "Epoch: 186, Loss: 0.4317\n",
            "Epoch: 187, Loss: 0.4911\n",
            "Epoch: 188, Loss: 0.3369\n",
            "Epoch: 189, Loss: 0.4945\n",
            "Epoch: 190, Loss: 0.3912\n",
            "Epoch: 191, Loss: 0.3824\n",
            "Epoch: 192, Loss: 0.3479\n",
            "Epoch: 193, Loss: 0.3798\n",
            "Epoch: 194, Loss: 0.3799\n",
            "Epoch: 195, Loss: 0.4015\n",
            "Epoch: 196, Loss: 0.3615\n",
            "Epoch: 197, Loss: 0.3985\n",
            "Epoch: 198, Loss: 0.4664\n",
            "Epoch: 199, Loss: 0.3714\n",
            "Epoch: 200, Loss: 0.3810\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "After training the model, we can call the test function to see how well our model performs on unseen labels. Here, we are interested in the accuracy of the model, i.e., the ratio of correctly classified nodes:"
      ],
      "metadata": {
        "id": "Dxex73fXXDXY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "test_acc = test()\n",
        "print(f'Test Accuracy: {test_acc:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vxdjUewgXDqY",
        "outputId": "95e4affb-cfe0-4f26-dc3c-879b5a0a8f5e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Accuracy: 0.5900\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "As one can see, our MLP performs rather bad with only about 59% test accuracy.\n",
        "\n",
        "**But why does the MLP do not perform better?**\n",
        "\n",
        "The main reason for that is that this model suffers from heavy overfitting due to only having access to a **small amount of training nodes**, and therefore generalizes poorly to unseen node representations.\n",
        "\n",
        "It also fails to incorporate an important bias into the model: **Cited papers are very likely related to the category of a document**.\n",
        "That is exactly where Graph Neural Networks come into play and can help to boost the performance of our model.\n",
        "\n"
      ],
      "metadata": {
        "id": "Qxr5-YOAXN7l"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Training a Graph Convolutional Neural Network (GCN)\n"
      ],
      "metadata": {
        "id": "XLJ1DJEd8zdk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn.functional as F\n",
        "from torch_geometric.nn import GCNConv\n",
        "\n",
        "class GCN(torch.nn.Module):\n",
        "    def __init__(self, hidden_channels):\n",
        "        super().__init__()\n",
        "        torch.manual_seed(1234567)\n",
        "        self.conv1 = GCNConv(dataset.num_features, hidden_channels)\n",
        "        self.conv2 = GCNConv(hidden_channels, dataset.num_classes)\n",
        "\n",
        "    def forward(self, x, edge_index):\n",
        "        x = self.conv1(x, edge_index)\n",
        "        x = x.relu()\n",
        "        x = F.dropout(x, p=0.5, training=self.training)\n",
        "        x = self.conv2(x, edge_index)\n",
        "        return x\n",
        "\n",
        "model = GCN(hidden_channels=16)\n",
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EliqmrWR_L52",
        "outputId": "ef9257fc-56f4-45b6-d239-fff12426abf1"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "GCN(\n",
            "  (conv1): GCNConv(1433, 16)\n",
            "  (conv2): GCNConv(16, 7)\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Define the training and validation functions for the GCN model"
      ],
      "metadata": {
        "id": "TXEoAV0zBSWu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model = GCN(hidden_channels=16)\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=0.01, weight_decay=5e-4)\n",
        "criterion = torch.nn.CrossEntropyLoss()\n",
        "\n",
        "def train():\n",
        "      model.train()\n",
        "      optimizer.zero_grad()  # Clear gradients.\n",
        "      out = model(data.x, data.edge_index)  # Perform a single forward pass.\n",
        "      loss = criterion(out[data.train_mask], data.y[data.train_mask])  # Compute the loss solely based on the training nodes.\n",
        "      loss.backward()  # Derive gradients.\n",
        "      optimizer.step()  # Update parameters based on gradients.\n",
        "      return loss\n",
        "\n",
        "def test():\n",
        "      model.eval()\n",
        "      out = model(data.x, data.edge_index)\n",
        "      pred = out.argmax(dim=1)  # Use the class with highest probability.\n",
        "      test_correct = pred[data.test_mask] == data.y[data.test_mask]  # Check against ground-truth labels.\n",
        "      test_acc = int(test_correct.sum()) / int(data.test_mask.sum())  # Derive ratio of correct predictions.\n",
        "      return test_acc\n"
      ],
      "metadata": {
        "id": "APBRjtp-BKsC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Train the model"
      ],
      "metadata": {
        "id": "pHDAgMoRB9sl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for epoch in range(1, 101):\n",
        "    loss = train()\n",
        "    print(f'Epoch: {epoch:03d}, Loss: {loss:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CcX9LX__B_WB",
        "outputId": "f0712962-32d0-4111-f6cd-67e2459675a4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch: 001, Loss: 1.9463\n",
            "Epoch: 002, Loss: 1.9409\n",
            "Epoch: 003, Loss: 1.9343\n",
            "Epoch: 004, Loss: 1.9275\n",
            "Epoch: 005, Loss: 1.9181\n",
            "Epoch: 006, Loss: 1.9086\n",
            "Epoch: 007, Loss: 1.9015\n",
            "Epoch: 008, Loss: 1.8933\n",
            "Epoch: 009, Loss: 1.8808\n",
            "Epoch: 010, Loss: 1.8685\n",
            "Epoch: 011, Loss: 1.8598\n",
            "Epoch: 012, Loss: 1.8482\n",
            "Epoch: 013, Loss: 1.8290\n",
            "Epoch: 014, Loss: 1.8233\n",
            "Epoch: 015, Loss: 1.8057\n",
            "Epoch: 016, Loss: 1.7966\n",
            "Epoch: 017, Loss: 1.7825\n",
            "Epoch: 018, Loss: 1.7617\n",
            "Epoch: 019, Loss: 1.7491\n",
            "Epoch: 020, Loss: 1.7310\n",
            "Epoch: 021, Loss: 1.7147\n",
            "Epoch: 022, Loss: 1.7056\n",
            "Epoch: 023, Loss: 1.6954\n",
            "Epoch: 024, Loss: 1.6697\n",
            "Epoch: 025, Loss: 1.6538\n",
            "Epoch: 026, Loss: 1.6312\n",
            "Epoch: 027, Loss: 1.6161\n",
            "Epoch: 028, Loss: 1.5899\n",
            "Epoch: 029, Loss: 1.5711\n",
            "Epoch: 030, Loss: 1.5576\n",
            "Epoch: 031, Loss: 1.5393\n",
            "Epoch: 032, Loss: 1.5137\n",
            "Epoch: 033, Loss: 1.4948\n",
            "Epoch: 034, Loss: 1.4913\n",
            "Epoch: 035, Loss: 1.4698\n",
            "Epoch: 036, Loss: 1.3998\n",
            "Epoch: 037, Loss: 1.4041\n",
            "Epoch: 038, Loss: 1.3761\n",
            "Epoch: 039, Loss: 1.3631\n",
            "Epoch: 040, Loss: 1.3258\n",
            "Epoch: 041, Loss: 1.3030\n",
            "Epoch: 042, Loss: 1.3119\n",
            "Epoch: 043, Loss: 1.2519\n",
            "Epoch: 044, Loss: 1.2530\n",
            "Epoch: 045, Loss: 1.2492\n",
            "Epoch: 046, Loss: 1.2205\n",
            "Epoch: 047, Loss: 1.2037\n",
            "Epoch: 048, Loss: 1.1571\n",
            "Epoch: 049, Loss: 1.1700\n",
            "Epoch: 050, Loss: 1.1296\n",
            "Epoch: 051, Loss: 1.0860\n",
            "Epoch: 052, Loss: 1.1080\n",
            "Epoch: 053, Loss: 1.0564\n",
            "Epoch: 054, Loss: 1.0157\n",
            "Epoch: 055, Loss: 1.0362\n",
            "Epoch: 056, Loss: 1.0328\n",
            "Epoch: 057, Loss: 1.0058\n",
            "Epoch: 058, Loss: 0.9865\n",
            "Epoch: 059, Loss: 0.9667\n",
            "Epoch: 060, Loss: 0.9741\n",
            "Epoch: 061, Loss: 0.9769\n",
            "Epoch: 062, Loss: 0.9122\n",
            "Epoch: 063, Loss: 0.8993\n",
            "Epoch: 064, Loss: 0.8769\n",
            "Epoch: 065, Loss: 0.8575\n",
            "Epoch: 066, Loss: 0.8897\n",
            "Epoch: 067, Loss: 0.8312\n",
            "Epoch: 068, Loss: 0.8262\n",
            "Epoch: 069, Loss: 0.8511\n",
            "Epoch: 070, Loss: 0.7711\n",
            "Epoch: 071, Loss: 0.8012\n",
            "Epoch: 072, Loss: 0.7529\n",
            "Epoch: 073, Loss: 0.7525\n",
            "Epoch: 074, Loss: 0.7689\n",
            "Epoch: 075, Loss: 0.7553\n",
            "Epoch: 076, Loss: 0.7032\n",
            "Epoch: 077, Loss: 0.7326\n",
            "Epoch: 078, Loss: 0.7122\n",
            "Epoch: 079, Loss: 0.7090\n",
            "Epoch: 080, Loss: 0.6755\n",
            "Epoch: 081, Loss: 0.6666\n",
            "Epoch: 082, Loss: 0.6679\n",
            "Epoch: 083, Loss: 0.7037\n",
            "Epoch: 084, Loss: 0.6752\n",
            "Epoch: 085, Loss: 0.6266\n",
            "Epoch: 086, Loss: 0.6564\n",
            "Epoch: 087, Loss: 0.6266\n",
            "Epoch: 088, Loss: 0.6411\n",
            "Epoch: 089, Loss: 0.6226\n",
            "Epoch: 090, Loss: 0.6535\n",
            "Epoch: 091, Loss: 0.6317\n",
            "Epoch: 092, Loss: 0.5741\n",
            "Epoch: 093, Loss: 0.5572\n",
            "Epoch: 094, Loss: 0.5710\n",
            "Epoch: 095, Loss: 0.5816\n",
            "Epoch: 096, Loss: 0.5745\n",
            "Epoch: 097, Loss: 0.5547\n",
            "Epoch: 098, Loss: 0.5989\n",
            "Epoch: 099, Loss: 0.6021\n",
            "Epoch: 100, Loss: 0.5799\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Predict on the test set"
      ],
      "metadata": {
        "id": "w-AjlfIq_g3E"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "test_acc = test()\n",
        "print(f'Test Accuracy: {test_acc:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pd3ZfnEnBsLB",
        "outputId": "ac2461d3-99a4-4613-b9a4-3afc0917e76b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Accuracy: 0.8150\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Graph Classification with GCN"
      ],
      "metadata": {
        "id": "kaMlqDMOLalD"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "The most common task for graph classification is **molecular property prediction**, in which molecules are represented as graphs, and the task may be to infer whether a molecule inhibits HIV virus replication or not.\n",
        "\n",
        "The TU Dortmund University has collected a wide range of different graph classification datasets, known as the [**TUDatasets**](https://chrsmrrs.github.io/datasets/), which are also accessible via [`torch_geometric.datasets.TUDataset`](https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html#torch_geometric.datasets.TUDataset) in PyTorch Geometric.\n",
        "Let's load and inspect one of the smaller ones, the **MUTAG dataset**.\n",
        "\n",
        "![](https://github.com/Lightning-AI/lightning-tutorials/raw/main/course_UvA-DL/06-graph-neural-networks/molecule_graph.svg)"
      ],
      "metadata": {
        "id": "2F5SGY39MYZe"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "from torch_geometric.datasets import TUDataset\n",
        "\n",
        "dataset = TUDataset(root='data/TUDataset', name='MUTAG')\n",
        "\n",
        "print()\n",
        "print(f'Dataset: {dataset}:')\n",
        "print('====================')\n",
        "print(f'Number of graphs: {len(dataset)}')\n",
        "print(f'Number of features: {dataset.num_features}')\n",
        "print(f'Number of classes: {dataset.num_classes}')\n",
        "\n",
        "data = dataset[0]  # Get the first graph object.\n",
        "\n",
        "print()\n",
        "print(data)\n",
        "print('=============================================================')\n",
        "\n",
        "# Gather some statistics about the first graph.\n",
        "print(f'Number of nodes: {data.num_nodes}')\n",
        "print(f'Number of edges: {data.num_edges}')\n",
        "print(f'Average node degree: {data.num_edges / data.num_nodes:.2f}')\n",
        "print(f'Has isolated nodes: {data.has_isolated_nodes()}')\n",
        "print(f'Has self-loops: {data.has_self_loops()}')\n",
        "print(f'Is undirected: {data.is_undirected()}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-BqSjtwLMY5e",
        "outputId": "c99a910b-a284-49fd-f34e-966849fa8e9f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Downloading https://www.chrsmrrs.com/graphkerneldatasets/MUTAG.zip\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Dataset: MUTAG(188):\n",
            "====================\n",
            "Number of graphs: 188\n",
            "Number of features: 7\n",
            "Number of classes: 2\n",
            "\n",
            "Data(edge_index=[2, 38], x=[17, 7], edge_attr=[38, 4], y=[1])\n",
            "=============================================================\n",
            "Number of nodes: 17\n",
            "Number of edges: 38\n",
            "Average node degree: 2.24\n",
            "Has isolated nodes: False\n",
            "Has self-loops: False\n",
            "Is undirected: True\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Processing...\n",
            "Done!\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Graph visualization"
      ],
      "metadata": {
        "id": "syX_z7OZOe6A"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "pip install pyvis"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JF42jOeDPDAI",
        "outputId": "f3969fb4-f121-4a47-d0df-77a219670179"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting pyvis\n",
            "  Downloading pyvis-0.3.2-py3-none-any.whl (756 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m756.0/756.0 kB\u001b[0m \u001b[31m6.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: ipython>=5.3.0 in /usr/local/lib/python3.10/dist-packages (from pyvis) (7.34.0)\n",
            "Requirement already satisfied: jinja2>=2.9.6 in /usr/local/lib/python3.10/dist-packages (from pyvis) (3.1.3)\n",
            "Requirement already satisfied: jsonpickle>=1.4.1 in /usr/local/lib/python3.10/dist-packages (from pyvis) (3.0.3)\n",
            "Requirement already satisfied: networkx>=1.11 in /usr/local/lib/python3.10/dist-packages (from pyvis) (3.2.1)\n",
            "Requirement already satisfied: setuptools>=18.5 in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (67.7.2)\n",
            "Collecting jedi>=0.16 (from ipython>=5.3.0->pyvis)\n",
            "  Downloading jedi-0.19.1-py2.py3-none-any.whl (1.6 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.6/1.6 MB\u001b[0m \u001b[31m43.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: decorator in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (4.4.2)\n",
            "Requirement already satisfied: pickleshare in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (0.7.5)\n",
            "Requirement already satisfied: traitlets>=4.2 in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (5.7.1)\n",
            "Requirement already satisfied: prompt-toolkit!=3.0.0,!=3.0.1,<3.1.0,>=2.0.0 in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (3.0.43)\n",
            "Requirement already satisfied: pygments in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (2.16.1)\n",
            "Requirement already satisfied: backcall in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (0.2.0)\n",
            "Requirement already satisfied: matplotlib-inline in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (0.1.6)\n",
            "Requirement already satisfied: pexpect>4.3 in /usr/local/lib/python3.10/dist-packages (from ipython>=5.3.0->pyvis) (4.9.0)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2>=2.9.6->pyvis) (2.1.5)\n",
            "Requirement already satisfied: parso<0.9.0,>=0.8.3 in /usr/local/lib/python3.10/dist-packages (from jedi>=0.16->ipython>=5.3.0->pyvis) (0.8.3)\n",
            "Requirement already satisfied: ptyprocess>=0.5 in /usr/local/lib/python3.10/dist-packages (from pexpect>4.3->ipython>=5.3.0->pyvis) (0.7.0)\n",
            "Requirement already satisfied: wcwidth in /usr/local/lib/python3.10/dist-packages (from prompt-toolkit!=3.0.0,!=3.0.1,<3.1.0,>=2.0.0->ipython>=5.3.0->pyvis) (0.2.13)\n",
            "Installing collected packages: jedi, pyvis\n",
            "Successfully installed jedi-0.19.1 pyvis-0.3.2\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import networkx as nx\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Convert the PyG Data object to a NetworkX graph\n",
        "G = nx.Graph()\n",
        "G.add_nodes_from(range(data.num_nodes))\n",
        "edge_index = data.edge_index.t().tolist()\n",
        "G.add_edges_from(edge_index)\n",
        "\n",
        "# Visualize the graph\n",
        "plt.figure(figsize=(8, 6))\n",
        "pos = nx.spring_layout(G, seed=42)  # Positions for all nodes\n",
        "nx.draw(G, pos, with_labels=True, node_color='skyblue', node_size=100, edge_color='gray', linewidths=0.5)\n",
        "plt.title('Visualization of a Graph from the MUTAG Dataset')\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 659
        },
        "id": "wlALzI7jOHBJ",
        "outputId": "3030c6c3-2ada-4627-b615-bae6a72f2fc3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 800x600 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "This dataset provides **188 different graphs**, and the task is to classify each graph into **one out of two classes**.\n",
        "\n",
        "By inspecting the first graph object of the dataset, we can see that it comes with **17 nodes (with 7-dimensional feature vectors)** and **38 edges** (leading to an average node degree of 2.24).\n",
        "It also comes with exactly **one graph label** (`y=[1]`), and, in addition to previous datasets, provides addtional **4-dimensional edge features** (`edge_attr=[38, 4]`).\n",
        "However, for the sake of simplicity, we will not make use of those.\n",
        "\n",
        "PyTorch Geometric provides some useful utilities for working with graph datasets, *e.g.*, we can shuffle the dataset and use the first 150 graphs as training graphs, while using the remaining ones for testing:"
      ],
      "metadata": {
        "id": "3-qZWB0PMiBV"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "torch.manual_seed(12345)\n",
        "dataset = dataset.shuffle()\n",
        "\n",
        "train_dataset = dataset[:150]\n",
        "test_dataset = dataset[150:]\n",
        "\n",
        "print(f'Number of training graphs: {len(train_dataset)}')\n",
        "print(f'Number of test graphs: {len(test_dataset)}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "C0PLzpxrMlkM",
        "outputId": "18abec2d-e7e5-4573-ee19-aee76c90b822"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of training graphs: 150\n",
            "Number of test graphs: 38\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Mini-batching of graphs\n",
        "\n",
        "Since graphs in graph classification datasets are usually small, a good idea is to **batch the graphs** before inputting them into a Graph Neural Network to guarantee full GPU utilization.\n",
        "In the image or language domain, this procedure is typically achieved by **rescaling** or **padding** each example into a set of equally-sized shapes, and examples are then grouped in an additional dimension.\n",
        "The length of this dimension is then equal to the number of examples grouped in a mini-batch and is typically referred to as the `batch_size`.\n",
        "\n",
        "However, for GNNs the two approaches described above are either not feasible or may result in a lot of unnecessary memory consumption.\n",
        "Therefore, PyTorch Geometric opts for another approach to achieve parallelization across a number of examples. Here, adjacency matrices are stacked in a diagonal fashion (creating a giant graph that holds multiple isolated subgraphs), and node and target features are simply concatenated in the node dimension.\n",
        "\n",
        "![Screen Shot 2020-08-27 at 13.11.53.png](data:image/png;base64,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)"
      ],
      "metadata": {
        "id": "zlLNurn6Mkm8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from torch_geometric.loader import DataLoader\n",
        "\n",
        "train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)\n",
        "test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)\n",
        "\n",
        "for step, data in enumerate(train_loader):\n",
        "    print(f'Step {step + 1}:')\n",
        "    print('=======')\n",
        "    print(f'Number of graphs in the current batch: {data.num_graphs}')\n",
        "    print(data)\n",
        "    print()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5n40YXo6MtWL",
        "outputId": "3addf3e3-5e0f-4513-874f-83fc0a53822a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Step 1:\n",
            "=======\n",
            "Number of graphs in the current batch: 64\n",
            "DataBatch(edge_index=[2, 2636], x=[1188, 7], edge_attr=[2636, 4], y=[64], batch=[1188], ptr=[65])\n",
            "\n",
            "Step 2:\n",
            "=======\n",
            "Number of graphs in the current batch: 64\n",
            "DataBatch(edge_index=[2, 2506], x=[1139, 7], edge_attr=[2506, 4], y=[64], batch=[1139], ptr=[65])\n",
            "\n",
            "Step 3:\n",
            "=======\n",
            "Number of graphs in the current batch: 22\n",
            "DataBatch(edge_index=[2, 852], x=[387, 7], edge_attr=[852, 4], y=[22], batch=[387], ptr=[23])\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Training the GCN\n",
        "\n",
        "Training a GCN for graph classification usually follows a simple recipe:\n",
        "\n",
        "1. Embed each node by performing multiple rounds of message passing\n",
        "2. Aggregate node embeddings into a unified graph embedding (**readout layer**)\n",
        "3. Train a final classifier on the graph embedding"
      ],
      "metadata": {
        "id": "95_3v2y0NR7b"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from torch.nn import Linear\n",
        "import torch.nn.functional as F\n",
        "from torch_geometric.nn import GCNConv\n",
        "from torch_geometric.nn import global_mean_pool\n",
        "\n",
        "\n",
        "class GCN(torch.nn.Module):\n",
        "    def __init__(self, hidden_channels):\n",
        "        super(GCN, self).__init__()\n",
        "        torch.manual_seed(12345)\n",
        "        self.conv1 = GCNConv(dataset.num_node_features, hidden_channels)\n",
        "        self.conv2 = GCNConv(hidden_channels, hidden_channels)\n",
        "        self.conv3 = GCNConv(hidden_channels, hidden_channels)\n",
        "        self.lin = Linear(hidden_channels, dataset.num_classes)\n",
        "\n",
        "    def forward(self, x, edge_index, batch):\n",
        "        # 1. Obtain node embeddings\n",
        "        x = self.conv1(x, edge_index)\n",
        "        x = x.relu()\n",
        "        x = self.conv2(x, edge_index)\n",
        "        x = x.relu()\n",
        "        x = self.conv3(x, edge_index)\n",
        "\n",
        "        # 2. Readout layer\n",
        "        x = global_mean_pool(x, batch)  # [batch_size, hidden_channels]\n",
        "\n",
        "        # 3. Apply a final classifier\n",
        "        x = F.dropout(x, p=0.5, training=self.training)\n",
        "        x = self.lin(x)\n",
        "\n",
        "        return x\n",
        "\n",
        "model = GCN(hidden_channels=64)\n",
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YtegtVsTNRns",
        "outputId": "dc7082f6-37fd-4576-c7b3-ba67f2b532f2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "GCN(\n",
            "  (conv1): GCNConv(7, 64)\n",
            "  (conv2): GCNConv(64, 64)\n",
            "  (conv3): GCNConv(64, 64)\n",
            "  (lin): Linear(in_features=64, out_features=2, bias=True)\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let's train our network for a few epochs to see how well it performs on the training as well as test set."
      ],
      "metadata": {
        "id": "eyxckreONSz9"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model = GCN(hidden_channels=64)\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=0.01)\n",
        "criterion = torch.nn.CrossEntropyLoss()\n",
        "\n",
        "def train():\n",
        "    model.train()\n",
        "\n",
        "    for data in train_loader:  # Iterate in batches over the training dataset.\n",
        "         out = model(data.x, data.edge_index, data.batch)  # Perform a single forward pass.\n",
        "         loss = criterion(out, data.y)  # Compute the loss.\n",
        "         loss.backward()  # Derive gradients.\n",
        "         optimizer.step()  # Update parameters based on gradients.\n",
        "         optimizer.zero_grad()  # Clear gradients.\n",
        "\n",
        "def test(loader):\n",
        "     model.eval()\n",
        "\n",
        "     correct = 0\n",
        "     for data in loader:  # Iterate in batches over the training/test dataset.\n",
        "         out = model(data.x, data.edge_index, data.batch)\n",
        "         pred = out.argmax(dim=1)  # Use the class with highest probability.\n",
        "         correct += int((pred == data.y).sum())  # Check against ground-truth labels.\n",
        "     return correct / len(loader.dataset)  # Derive ratio of correct predictions.\n",
        "\n",
        "\n",
        "for epoch in range(1, 171):\n",
        "    train()\n",
        "    train_acc = test(train_loader)\n",
        "    print(f'Epoch: {epoch:03d}, Train Acc: {train_acc:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_3wtLT9nNIvs",
        "outputId": "db92efd2-5244-41a2-82a4-e4dc4764bf09"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch: 001, Train Acc: 0.6467\n",
            "Epoch: 002, Train Acc: 0.6467\n",
            "Epoch: 003, Train Acc: 0.6467\n",
            "Epoch: 004, Train Acc: 0.6467\n",
            "Epoch: 005, Train Acc: 0.6467\n",
            "Epoch: 006, Train Acc: 0.6867\n",
            "Epoch: 007, Train Acc: 0.7133\n",
            "Epoch: 008, Train Acc: 0.7467\n",
            "Epoch: 009, Train Acc: 0.7133\n",
            "Epoch: 010, Train Acc: 0.7200\n",
            "Epoch: 011, Train Acc: 0.7267\n",
            "Epoch: 012, Train Acc: 0.7067\n",
            "Epoch: 013, Train Acc: 0.7200\n",
            "Epoch: 014, Train Acc: 0.7333\n",
            "Epoch: 015, Train Acc: 0.7200\n",
            "Epoch: 016, Train Acc: 0.7133\n",
            "Epoch: 017, Train Acc: 0.7333\n",
            "Epoch: 018, Train Acc: 0.7600\n",
            "Epoch: 019, Train Acc: 0.7200\n",
            "Epoch: 020, Train Acc: 0.7267\n",
            "Epoch: 021, Train Acc: 0.7533\n",
            "Epoch: 022, Train Acc: 0.7533\n",
            "Epoch: 023, Train Acc: 0.7267\n",
            "Epoch: 024, Train Acc: 0.7400\n",
            "Epoch: 025, Train Acc: 0.7533\n",
            "Epoch: 026, Train Acc: 0.7533\n",
            "Epoch: 027, Train Acc: 0.7400\n",
            "Epoch: 028, Train Acc: 0.7333\n",
            "Epoch: 029, Train Acc: 0.7467\n",
            "Epoch: 030, Train Acc: 0.7467\n",
            "Epoch: 031, Train Acc: 0.7600\n",
            "Epoch: 032, Train Acc: 0.7667\n",
            "Epoch: 033, Train Acc: 0.7600\n",
            "Epoch: 034, Train Acc: 0.7600\n",
            "Epoch: 035, Train Acc: 0.7467\n",
            "Epoch: 036, Train Acc: 0.7600\n",
            "Epoch: 037, Train Acc: 0.7600\n",
            "Epoch: 038, Train Acc: 0.7600\n",
            "Epoch: 039, Train Acc: 0.7533\n",
            "Epoch: 040, Train Acc: 0.7533\n",
            "Epoch: 041, Train Acc: 0.7600\n",
            "Epoch: 042, Train Acc: 0.7733\n",
            "Epoch: 043, Train Acc: 0.7667\n",
            "Epoch: 044, Train Acc: 0.7667\n",
            "Epoch: 045, Train Acc: 0.7667\n",
            "Epoch: 046, Train Acc: 0.7667\n",
            "Epoch: 047, Train Acc: 0.7667\n",
            "Epoch: 048, Train Acc: 0.7667\n",
            "Epoch: 049, Train Acc: 0.7533\n",
            "Epoch: 050, Train Acc: 0.7733\n",
            "Epoch: 051, Train Acc: 0.7733\n",
            "Epoch: 052, Train Acc: 0.7667\n",
            "Epoch: 053, Train Acc: 0.7667\n",
            "Epoch: 054, Train Acc: 0.7733\n",
            "Epoch: 055, Train Acc: 0.7667\n",
            "Epoch: 056, Train Acc: 0.7667\n",
            "Epoch: 057, Train Acc: 0.7667\n",
            "Epoch: 058, Train Acc: 0.7800\n",
            "Epoch: 059, Train Acc: 0.7667\n",
            "Epoch: 060, Train Acc: 0.7667\n",
            "Epoch: 061, Train Acc: 0.7733\n",
            "Epoch: 062, Train Acc: 0.7800\n",
            "Epoch: 063, Train Acc: 0.7400\n",
            "Epoch: 064, Train Acc: 0.7667\n",
            "Epoch: 065, Train Acc: 0.7800\n",
            "Epoch: 066, Train Acc: 0.7667\n",
            "Epoch: 067, Train Acc: 0.7533\n",
            "Epoch: 068, Train Acc: 0.7867\n",
            "Epoch: 069, Train Acc: 0.7800\n",
            "Epoch: 070, Train Acc: 0.7667\n",
            "Epoch: 071, Train Acc: 0.7667\n",
            "Epoch: 072, Train Acc: 0.7800\n",
            "Epoch: 073, Train Acc: 0.7800\n",
            "Epoch: 074, Train Acc: 0.7667\n",
            "Epoch: 075, Train Acc: 0.7800\n",
            "Epoch: 076, Train Acc: 0.7867\n",
            "Epoch: 077, Train Acc: 0.7667\n",
            "Epoch: 078, Train Acc: 0.7600\n",
            "Epoch: 079, Train Acc: 0.7733\n",
            "Epoch: 080, Train Acc: 0.8067\n",
            "Epoch: 081, Train Acc: 0.7800\n",
            "Epoch: 082, Train Acc: 0.7733\n",
            "Epoch: 083, Train Acc: 0.7733\n",
            "Epoch: 084, Train Acc: 0.7800\n",
            "Epoch: 085, Train Acc: 0.7733\n",
            "Epoch: 086, Train Acc: 0.7800\n",
            "Epoch: 087, Train Acc: 0.7733\n",
            "Epoch: 088, Train Acc: 0.7933\n",
            "Epoch: 089, Train Acc: 0.7733\n",
            "Epoch: 090, Train Acc: 0.7800\n",
            "Epoch: 091, Train Acc: 0.7667\n",
            "Epoch: 092, Train Acc: 0.7667\n",
            "Epoch: 093, Train Acc: 0.7667\n",
            "Epoch: 094, Train Acc: 0.7733\n",
            "Epoch: 095, Train Acc: 0.7733\n",
            "Epoch: 096, Train Acc: 0.8133\n",
            "Epoch: 097, Train Acc: 0.7800\n",
            "Epoch: 098, Train Acc: 0.7933\n",
            "Epoch: 099, Train Acc: 0.8000\n",
            "Epoch: 100, Train Acc: 0.7733\n",
            "Epoch: 101, Train Acc: 0.8000\n",
            "Epoch: 102, Train Acc: 0.7867\n",
            "Epoch: 103, Train Acc: 0.7867\n",
            "Epoch: 104, Train Acc: 0.7733\n",
            "Epoch: 105, Train Acc: 0.8000\n",
            "Epoch: 106, Train Acc: 0.7933\n",
            "Epoch: 107, Train Acc: 0.8000\n",
            "Epoch: 108, Train Acc: 0.7733\n",
            "Epoch: 109, Train Acc: 0.7800\n",
            "Epoch: 110, Train Acc: 0.7933\n",
            "Epoch: 111, Train Acc: 0.7867\n",
            "Epoch: 112, Train Acc: 0.8000\n",
            "Epoch: 113, Train Acc: 0.8000\n",
            "Epoch: 114, Train Acc: 0.7867\n",
            "Epoch: 115, Train Acc: 0.8000\n",
            "Epoch: 116, Train Acc: 0.7733\n",
            "Epoch: 117, Train Acc: 0.8133\n",
            "Epoch: 118, Train Acc: 0.7933\n",
            "Epoch: 119, Train Acc: 0.8000\n",
            "Epoch: 120, Train Acc: 0.8133\n",
            "Epoch: 121, Train Acc: 0.7867\n",
            "Epoch: 122, Train Acc: 0.7667\n",
            "Epoch: 123, Train Acc: 0.8067\n",
            "Epoch: 124, Train Acc: 0.8067\n",
            "Epoch: 125, Train Acc: 0.7867\n",
            "Epoch: 126, Train Acc: 0.8133\n",
            "Epoch: 127, Train Acc: 0.7933\n",
            "Epoch: 128, Train Acc: 0.7933\n",
            "Epoch: 129, Train Acc: 0.8000\n",
            "Epoch: 130, Train Acc: 0.8067\n",
            "Epoch: 131, Train Acc: 0.7867\n",
            "Epoch: 132, Train Acc: 0.8133\n",
            "Epoch: 133, Train Acc: 0.8200\n",
            "Epoch: 134, Train Acc: 0.8133\n",
            "Epoch: 135, Train Acc: 0.7867\n",
            "Epoch: 136, Train Acc: 0.8267\n",
            "Epoch: 137, Train Acc: 0.8067\n",
            "Epoch: 138, Train Acc: 0.8400\n",
            "Epoch: 139, Train Acc: 0.8200\n",
            "Epoch: 140, Train Acc: 0.8333\n",
            "Epoch: 141, Train Acc: 0.8067\n",
            "Epoch: 142, Train Acc: 0.8200\n",
            "Epoch: 143, Train Acc: 0.8267\n",
            "Epoch: 144, Train Acc: 0.8133\n",
            "Epoch: 145, Train Acc: 0.8267\n",
            "Epoch: 146, Train Acc: 0.8333\n",
            "Epoch: 147, Train Acc: 0.8133\n",
            "Epoch: 148, Train Acc: 0.8267\n",
            "Epoch: 149, Train Acc: 0.8267\n",
            "Epoch: 150, Train Acc: 0.8267\n",
            "Epoch: 151, Train Acc: 0.8400\n",
            "Epoch: 152, Train Acc: 0.8467\n",
            "Epoch: 153, Train Acc: 0.8400\n",
            "Epoch: 154, Train Acc: 0.8400\n",
            "Epoch: 155, Train Acc: 0.8333\n",
            "Epoch: 156, Train Acc: 0.8267\n",
            "Epoch: 157, Train Acc: 0.7867\n",
            "Epoch: 158, Train Acc: 0.8200\n",
            "Epoch: 159, Train Acc: 0.8533\n",
            "Epoch: 160, Train Acc: 0.8467\n",
            "Epoch: 161, Train Acc: 0.8533\n",
            "Epoch: 162, Train Acc: 0.8467\n",
            "Epoch: 163, Train Acc: 0.8133\n",
            "Epoch: 164, Train Acc: 0.8400\n",
            "Epoch: 165, Train Acc: 0.8267\n",
            "Epoch: 166, Train Acc: 0.8400\n",
            "Epoch: 167, Train Acc: 0.8000\n",
            "Epoch: 168, Train Acc: 0.8533\n",
            "Epoch: 169, Train Acc: 0.8067\n",
            "Epoch: 170, Train Acc: 0.8333\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "test_acc = test(test_loader)\n",
        "print(f'Test Acc: {test_acc:.4f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2OBT2YsdncJr",
        "outputId": "9a8c3d85-c744-448b-a040-a01c0700aac6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Acc: 0.7895\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# HOMEWORK 7"
      ],
      "metadata": {
        "id": "vzxGLid4PpzC"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Perform node classification on the PubMed dataset to classify scientific publications into their respective categories. The PubMed dataset is a citation network among scientific publications. It consists of articles related to diabetes, with each article belonging to one of three categories: \"Diabetes Mellitus Type 1\" (class 0), \"Diabetes Mellitus Type 2\" (class 1), and \"Diabetes Mellitus\" (class 2)."
      ],
      "metadata": {
        "id": "87d53_SXSKn8"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Install required packages.\n"
      ],
      "metadata": {
        "id": "4ODQgEFgpBKk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import os\n",
        "import torch\n",
        "os.environ['TORCH'] = torch.__version__\n",
        "print(torch.__version__)\n",
        "\n",
        "!pip install -q torch-scatter -f https://data.pyg.org/whl/torch-${TORCH}.html\n",
        "!pip install -q torch-sparse -f https://data.pyg.org/whl/torch-${TORCH}.html\n",
        "!pip install -q git+https://github.com/pyg-team/pytorch_geometric.git"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jNcfcfjzpFwk",
        "outputId": "21eb644f-939b-4959-c8cd-95990a3f4b97"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2.2.1+cu121\n",
            "  Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n",
            "  Getting requirements to build wheel ... \u001b[?25l\u001b[?25hdone\n",
            "  Preparing metadata (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Load the dataset"
      ],
      "metadata": {
        "id": "nw-2CecKodxM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from torch_geometric.datasets import Planetoid\n",
        "\n",
        "# Load the Cora dataset\n",
        "dataset = Planetoid(root='data/Planetoid', name='PubMed')\n",
        "\n",
        "# Print dataset information\n",
        "print()\n",
        "print(f'Dataset: {dataset}:')\n",
        "print('======================')\n",
        "print(f'Number of graphs: {len(dataset)}')\n",
        "print(f'Number of features: {dataset.num_features}')\n",
        "print(f'Number of classes: {dataset.num_classes}')\n",
        "\n",
        "data = dataset[0]  # Get the first graph object.\n",
        "\n",
        "print()\n",
        "print(data)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5MxCzANaoiP1",
        "outputId": "a9998b32-e3a1-4d70-ad46-5546d4fe14ae"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.x\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.tx\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.allx\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.y\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.ty\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.ally\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.graph\n",
            "Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.pubmed.test.index\n",
            "Processing...\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Dataset: PubMed():\n",
            "======================\n",
            "Number of graphs: 1\n",
            "Number of features: 500\n",
            "Number of classes: 3\n",
            "\n",
            "Data(x=[19717, 500], edge_index=[2, 88648], y=[19717], train_mask=[19717], val_mask=[19717], test_mask=[19717])\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Done!\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Define a GCN network\n"
      ],
      "metadata": {
        "id": "DDez15KJseK_"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "Jq4PCKNMsjR2"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Visualize the node embeddings of the untrained GCN network using PCA"
      ],
      "metadata": {
        "id": "y3tk5298s2ue"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "kgxX95urtJgt"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Perform a grid search hyperparameter tuning for the following parameters\n",
        "- Number of Layers\n",
        "- Hidden Dimension Size\n",
        "- Learning Rate\n",
        "- Dropout Rate\n",
        "- Batch Size"
      ],
      "metadata": {
        "id": "pYX_Vp-Qt-Fr"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "vsNmuyxCv-74"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        " # Train the GCN model with the optimal hyperparameters and select the best model based on the highest validation performance"
      ],
      "metadata": {
        "id": "fp5rX-NcslFe"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "JRFMZwWVslvW"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Visualize the output embeddings of your trained model. What do you observe?"
      ],
      "metadata": {
        "id": "VFDb77Y0stAe"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "IMaMXYvnstf2"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Evaluate the trained model on the test set and report the classification accuracy"
      ],
      "metadata": {
        "id": "22G1yJext1ev"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "96LWgh5tt9gd"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}