Weekly Talk by Helen Zhang, Head, UA GIDP Statistics: Dynamic Supervised Principal Component Analysis for High-Dimensional Classification

Available in person and via zoom (see email for link)

Image

Abstract

High-dimensional data classification is challenged by distributions that shift dynamically over time, making static subspace definitions and decision boundaries inadequate. We propose a novel framework for dynamic classification in high-dimensional spaces, designed to accommodate evolving class distributions across time or other index variables. The framework employs a supervised dimension reduction technique based on kernel smoothing to identify an optimal subspace and construct adaptive classification boundaries that respond to distributional changes. We develop theory and computational algorithms for both linear and quadratic discriminant analysis, and illustrate effectiveness of the proposed approach through simulation studies and real data applications.

Bio

Helen Zhang, Professor of Mathematics, and Head, U of A GIDP Statistics