Speaker: Thomas Strohmer (UC Davis, Math)
Title: A Rigorous Framework for Data Clustering
Abstract: Organizing data into meaningful groups is one of the most fundamental
tasks in data analysis and machine learning. While spectral clustering has become one of the most popular clustering techniques, a rigorous and meaningful theoretical justification has still been elusive so far. I will propose a convex relaxation approach, which gives rise to a rigorous theoretical analysis of spectral clustering. We do this by deriving deterministic bounds of finding optimal graph cuts via a natural and intuitive proximity condition related to the spectrum of the graph Laplacian. Moreover, the proposed approach provides theoretical guarantees for community detection. I will discuss extensions and applications of our framework.
Roundtable Discussion will start immediately after this talk. The topics include: Challenges of high-dimensional data clustering; Relationship and interactions between CeDAR (IMPACT) and UCD4IDS (TRIPODS).
So, please attend both the seminar and the roundtable discussion!