Persistence in PyTorch

Primoz Skraba writes:
Announcing a persistent homology layer for PyTorch. The layer takes in either a simplicial complex or a point cloud, computes the persistence diagram and given an energy function automatically backpropogates. This allows for simple integration with other PyTorch modules as well as continuous optimization over persistence diagrams. Our goal is to make experimentation combining topology and deep networks easier and more accessible.

The module along with installation instructions and several examples can be found at https://github.com/bruel-gabrielsson/TopologyLayer

The companion paper describing the examples can be found at https://arxiv.org/abs/1905.12200

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.