Seminar Series - On the importance of low-dimensional representations for high-dimensional data

Speaker: Lawrence Carin, Duke University
Date: Friday, November 9, 2012
Time: 11:15am-12:30pm
Location: 2150 Torgersen

People are increasingly inundated with data, in the form of text, imagery, video and audio, with the internet an important delivery mechanism. A challenge concerns making sense of all of this data, and distilling it into useful information. In this talk we discuss the importance of low-dimensional models for representation of such multi-modality high-dimensional data. Low dimensionality is typically imposed in the form of sparsity and via low-rank representations. We discuss such models, methods of inference/computation, and their application to compact/meaningful representations of given high-dimensional data. We also discuss new compressive methods for data acquisition (compressive sensing).