Reproduce “Persistent reachability homology in machine learning applications” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luigi Caputi, Nicholas Meadows, Henri Riihimäki
Researchers from ETH Zurich and collaborators introduced Persistent Reachability Homology (PRH), a new topological method for directed graphs, and applied it to classify preictal and ictal brain states from EEG correlation networks. PRH achieved a peak classification accuracy of 82%, outperforming traditional Persistent Homology of Directed Flag Complexes (DPH) in 7 out of 8 comparisons.
Key insights
- Persistent Reachability Homology (PRH) captures distinct combinatorial and algebraic features of directed graphs more effectively than DPH, leveraging condensation to provide finer homological information.
- PRH offers computational advantages due to its condensation step, which drastically reduces the size of digraphs, making it faster to compute than DPH.
- Different homology theories emphasize different aspects of graph structure; DPH disproportionately relies on higher-degree homology (β2), while PRH focuses on lower-degree (β0, β1) features for classification.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.