Speaker: Lawrence Carin, Electrical Engineering, Duke University

Title: Robust Sparse Factor Analysis

When/where: Tuesday, February 15th, 4:15 PM, AKW 200

Abstract: A landmark-dependent hierarchical beta process
is developed as a prior for data with associated
covariates. The landmarks define local regions
in the covariate space where feature usages are
likely to be similar. The landmark locations are
learned, to which the data are linked through
normalized kernels. To demonstrate unique aspects
of the proposed model, we consider several
applications: (i) denoising of an image contaminated
by a superposition of Gaussian and spiky
noise, (ii) topic and spiky-keyword discovery
from a document corpora, (iii) analysis of hyperspectral data, and
(iv) foreground-background separation in video. State-of-the-art
performance is demonstrated, with efficient inference.