Open voetberg opened 1 year ago
Pytorch geometric has this already implemented https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.explain.algorithm.GNNExplainer.html
Needs to be tested on data, visuals iterated on
Example Visuals
Update to include the nexus node connections in the training -
Updated with filtering using topk hard thresholding
Applied this method to the actual network
Negative Fidelity metric is Miserable - so there are multiple courses of action
Updating the method to train for the top quartile instead of a topk. - Shows an improvement in characterization
Modified to create a secondary method that only works on each class (optimized results for each class separately). Results are as follows:
MIP
HIP
Michel
diffuse
Current updates - Targeting a class in the loss function for the mask optimizer, but not limiting the edges the filter can act on.
Showing fairly good positive and negative fidelity. Doing a full training run on the positive and negative examples. Loss curve and standard test sample below.
Further investigation
reference paper - https://arxiv.org/pdf/1903.03894.pdf
integrate into existing framework