search

ndey96 / GCNN-Explainability

Implementation of "Explainability Methods for Graph Convolutional Neural Networks" from HRL Laboratories
MIT License
77 stars 3 forks source link

#layers required for GCN explainability #4

Closed abhigoku10 closed 3 years ago

abhigoku10 commented 3 years ago

@ndey96 can this be applied to 3 layer gcn architecture with global pool ?? can we have two layer architecture before providing it to softmax function ?

ndey96 commented 3 years ago

Hi @abhigoku10 in models.py the model defined is a 3 layer GCNN, followed by global average pooling and a fully-connected output layer. If you would like to have a two layer architecture you can easily modify this file to do so :) Let me know if you have any questions