Topological Methods for Recommender Systems

Topology of preference networks for user–item recommendation.

A recent thread, with three Master’s students at GWU, applying topological data analysis to recommender systems. The user–item interaction graph is a bipartite network whose topology — connectedness, loops, voids in the preference space — carries signal that conventional matrix-factorisation and embedding pipelines discard. The work asks how persistence-based and Euler-characteristic descriptors of that network can sharpen recommendation, especially in cold-start and long-tail regimes.

Master’s students

• Alexander D. Silberman, M.S. Data Science, GWU
• Chinaza Belolisa, M.S. Data Science, GWU
• Madeline Bumpus, M.S. Data Science, GWU