The code for running the experiments of the "Capsule Neural Networks for Graph Classification using Explicit Tensorial Graph Representations" paper can be found in this repository.
The main scripts used are described here below:
- ExperimentsWithPatchyCapsules.ipynb: Experiments for finding good representations can be found in
- GraphClassifier.py: It builds and trains the Capsule Network graph classifier
- ConvNetPatchy.py: It builds and trains the CNN graph classifier
- PatchyTools/PatchyConverter: It Converts a graph in its Tensorial Representation.
All sample data/results stored on Dropbox as described here.