About feature

[huizhang1032]

How to output the feature of the last layer of the neural network of each graph in the list of graphs to be tested ?

[muhanzhang]

You can add a "-extract_features" to your command to save the test graphs' embeddings to "extracted_features_test.txt". See this line.

[huizhang1032]

OK.Thank you!

[huizhang1032]

If I input 100 graph objects to the model, the output feature is 10032. Does 132 here represent the embed of a graph object?

[muhanzhang]

Hi, if you input 100 graphs, and set the final graph embedding size to 32, then you will get 100 * (1 + 32) dimensional matrix in your output txt. The 1 additional dimension corresponds to the predicted labels of all input graphs.

[huizhang1032]

I got it.Thank you very much！