CODEWITHSUNDEEP
pythonpytorch-geometricgraph-neural-network
sumit kumar
sumit kumar

Reputation: 7

GraphConv vs GCNConv pytorch.geometric

I'm trying to compare 2 models GraphConv and GCNConv for my project. According to the tutorial provided by Pytorch GraphConv preserves central node information by omitting neighbourhood normalisation. The tutorial shows that GraphConv has improved the test accuracy to 82%, compared to 76%, in the case of the GCNConv model. However, in my case GraphConv increased the training accuracy from 89% (by GCNConv) to 93% but no change in test accuracy i.e. 88% (similar in GraphConv and GCNConv). I'm keeping the rest of the parameters the same, so what could be the reason(s)? Is there any way to increase the test accuracy? Thank you, below is the link to the Pytorch tutorial I mentioned here.

I tried to compare GraphConv and GCNConv, GraphConv is supposed to give me better test accuracy results, but it doesn't.

Upvotes: 0

Views: 1516

Answers (1)

Danii-Sh
Danii-Sh

Reputation: 455

Could you share your code for GraphConv? As the random parts of the code can alter the output I think running it a few times (if your code is working correclty) can show slightly different results, for eample, check my architecture for the network,

    self.conv1 = GraphConv(dataset.num_node_features, hidden_channels)
    self.conv2 = GraphConv(hidden_channels, hidden_channels)
    self.conv3 = GraphConv(hidden_channels, hidden_channels)
    self.lin = Linear(hidden_channels, dataset.num_classes)

This would yield a Test accuracy around 84 percent, As you can see in a sample run:

Epoch: 189, Train Acc: 0.8600, Test Acc: 0.7368
Epoch: 190, Train Acc: 0.9333, Test Acc: 0.7895
Epoch: 191, Train Acc: 0.9133, Test Acc: 0.7895
Epoch: 192, Train Acc: 0.9333, Test Acc: 0.7895
Epoch: 193, Train Acc: 0.9333, Test Acc: 0.7895
Epoch: 194, Train Acc: 0.9333, Test Acc: 0.7895
Epoch: 195, Train Acc: 0.9333, Test Acc: 0.8158
Epoch: 196, Train Acc: 0.9133, Test Acc: 0.7895
Epoch: 197, Train Acc: 0.9467, Test Acc: 0.8158
Epoch: 198, Train Acc: 0.9467, Test Acc: 0.8158
Epoch: 199, Train Acc: 0.9333, Test Acc: 0.7895
Epoch: 200, Train Acc: 0.9333, Test Acc: 0.8421

Upvotes: 0

Related Questions