/group_equivariant_capsules_pytorch

Pytorch implementation of Group Equivariant Capsule Networks

Primary LanguagePython

Group Equivariant Capsule Networks in PyTorch

Proof of concept Pytorch implementation of Group Equivariant Capsule Networks presented at NIPS 2018.

Required: pytorch_geometric

Install:

  1. Clone Repo
  2. python setup.py install

In repo dir: python examples/mnist.py to run the MNIST Experiment.

Implementations of capsules in group_capsules/nn/modules/pooling_capsule_layer.py

Implementations of sparse group convolution in group_capsules/nn/modules/group_conv_layer.py