Alternative convolution layers and clustering

[hkim716]

Hi Matt, SplineConv actually worked great, and it's pretty straightforward to understand compared to other type of graph convolutions. Thank you for your contribution on that one.
I'm trying to use other types of graph conv layer that outputs data.x.

class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = SplineConv(d.num_features, 32, dim=2, kernel_size=5)
        self.conv2 = SplineConv(32, 64, dim=2, kernel_size=5)
        self.fc1 = torch.nn.Linear(64, 128)
        self.fc2 = torch.nn.Linear(128, d.num_classes)

    def forward(self, data):
        data.x = self.conv1(data.x, data.edge_index, data.edge_attr)
        weight = normalized_cut_2d(data.edge_index, data.pos)
        cluster = graclus(data.edge_index, weight, data.x.size(0))
        data = max_pool(cluster, data, transform=transform)

        data.x = self.conv2(data.x, data.edge_index, data.edge_attr)
        weight = normalized_cut_2d(data.edge_index, data.pos)
        cluster = graclus(data.edge_index, weight, data.x.size(0))
        x, batch = max_pool_x(cluster, data.x, data.batch)

        x = global_mean_pool(x, batch)
        x = F.elu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        return F.log_softmax(self.fc2(x), dim=1)

In this example, SplineConv generates trainable parameters, and outputs new data.x using data.x, data.edge_index, data.edge_attr.
Is there any other graph convolution in PyG that takes input arguments similar to SplineConv, and outputs data.x?
And is there any other clustering algorithm in PyG that could be used in this example?

[rusty1s]

This is actually true for any GNN implemented in PyG. Every GNN will output new node features x, but not all can handle edge features as input. The cheatsheet should give you clarification on whether a GNN supports handling of edge features. Let me know if this is helpful to you.

[hkim716]

Thanks Matt this is very helpful! I will try to play with other type of convolutions and let you know if I have any question.