Neural Homology Theory

The learnability of different neural architectures can be characterized directly by computable measures of data complexity. Neural homology theory gives a complete topological description of the power of neural networks with applications in architecture selection and computational topology.

Papers, Presentations, Posters