Final Report: Geometric Factorization Tools for Community Mining
Abstract
Clustering nodes (users, entities) in a social network into different communities is an instrumental task in network analytics. Community detection and association is a core problem in machine learning, data mining,and signal processing. However, there are still many challenges remaining, and fundamental aspects that are poorly understood. Whether it is possible to identify the unknown underlying communities and pin down the association of nodes to these communities under realistic conditions (e.g., from a randomly or systematically sampled subset of links) is {\em a priori} unclear. Theory and scalable algorithms for extracting large quasi-cliques(densely connected communities) in a graph are sorely missing; and how to infer network typologies with limited prior knowledge has also been an open problem. This project developed a suite of principled analytical and computational tools to advance the state-of-art on several fronts in the broad area of network analytics and graph information retrieval. Its primary goal was the development, analysis, and validation of geometric matrix factorization tools for community mining.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 08, 2023
- Accession Number
- AD1225529
Entities
People
- Nikolaos Sidiropoulos
- Xiao Fu
- Zack W Almquist
Organizations
- University of Virginia