Node classification – predicting student interests

In this section, we’ll implement a GCN to predict student interests based on their features and connections in the social network. This task demonstrates how GNNs can leverage both node attributes and network structure to make predictions about individual nodes.

Let’s start by defining our GCN model:

class GCN(torch.nn.Module):
    def __init__(self, in_channels, hidden_channels, out_channels):
        super().__init__()
        self.conv1 = GCNConv(in_channels, hidden_channels)
        self.conv2 = GCNConv(hidden_channels, out_channels)
    def forward(self, x, edge_index):
        x = self.conv1(x, edge_index).relu()
        x = self.conv2(x, edge_index)
  ...