Tensorflow implementation of papers that use persistent homology features:
- Topological Autoencoders inspired by the authors' own PyTorch implementation
- Topologically Densified Distributions
In both cases, the interesting part is the calculation of 0D homology features (ie edges in the minimum spanning tree). This implementation includes a numpy implementation that runs on the CPU, and a pure Tensorflow tf.function implementation that runs in graph execution mode.
Topological Autoencoders, Michael Moor, Max Horn, Bastian Rieck, Karsten Borgwardt (ICML 2020)
Topologically Densified Distributions, Christoph Hofer, Florian Graf, Marc Niethammer, Roland Kwitt (ICML 2020)
Connectivity-Optimized Representation Learning via Persistent Homology, Christoph Hofer, Roland Kwitt, Marc Niethammer, Mandar Dixit (PMLR 97:2751-2760, 2019)