Skip to content

Back to Departmental Colloquium: Fall 2025

Departmental Colloquium


Date: Thursday, Nov 13, 2025

Time: 4:00PM - 5:00PM

Location: JWB 335


Kevin Miller

Brigham Young University

Title

Adaptive algorithms for data summarization

Abstract

Clustering, low-rank approximation, and nonnegative matrix factorization can be phrased as data summarization problems based on selecting a small number of prototypes from a large data set. We provide a unified formulation of these problems and many others. Natural algorithms for solving these problems include adaptive search, which is a greedy deterministic selection rule, and adaptive sampling, which is a randomized selection rule based on distances to existing prototypes. We evaluate the utility of the two approaches theoretically and empirically, pointing out when each approach is preferable.

Applied Mathematics Data Science and Machine Learning

Calendar file