A Chordless Cycle Filtration for Network Dimensionality Detection

Explore a novel topological data analysis approach for estimating the dimensionality of complex networks through a chordless cycle filtration scheme in this 58-minute conference talk. Learn how many real-world networks, from social to biological systems, exhibit structural patterns consistent with underlying hyperbolic geometry and discover why their latent space dimensionality is surprisingly low. Examine the innovative weighting scheme for graphs based on chordless cycles designed to estimate network dimensionality in a data-driven manner. Understand how the resulting topological descriptors can effectively predict network dimensionality using neural network architectures trained on synthetic graph databases. Gain insights into the intersection of algebraic topology, network analysis, and machine learning through research conducted in collaboration with experts in the field, demonstrating practical applications for complex network analysis across various domains.