Persistent homology of directed networks

Samir Chowdhury,F. Mémoli
DOI: https://doi.org/10.1109/ACSSC.2016.7868997
2016-11-01
Abstract:While persistent homology has been successfully used to provide topological summaries of point cloud data, the question of computing persistent homology of graphs or networks remains unclear. In particular, the existing literature does not provide a treatment of persistent homology for directed networks that is sensitive to asymmetry. We study a method for constructing simplicial complexes from weighted, directed networks that captures directionality information, and we are able to prove that the persistent homology of such complexes is stable with respect to a certain notion of network distance. We illustrate our construction on a database of simulated hippocampal networks.
What problem does this paper attempt to address?