Bayesian Topological Learning for Classifying the Structure of Biological Networks

Actin cytoskeleton networks generate local topological signatures due to the natural variations in number, size, and shape of holes networks. Persistent homology is a method that explores these properties data summarizes them as persistence diagrams. In this work, we analyze classify simulated actin filament by transforming into diagrams whose variability quantified via Bayesian framework on space The proposed generalized adopts an independent identically distributed cluster point process characterization relies substitution likelihood argument. This provides flexibility estimate posterior cardinality distribution points diagram their spatial simultaneously. We present closed form posteriors under assumption Gaussian mixture binomial for prior intensity respectively. Using calculation, finally, implement Bayes factor algorithm benchmark it against several state-of-the-art classification methods.