A geometric approach to modeling and analyzing dynamics in biomedical time series

Seminar: 
Applied Mathematics
Event time: 
Tuesday, November 19, 2019 - 4:00pm
Location: 
LOM 215
Speaker: 
John Malik
Speaker affiliation: 
Duke
Event description: 

Abstract:  Informed by physiology, the wave-shape oscillatory model for biomedical time series asserts that cycles lie on or near a low-dimensional manifold.

Recovering this manifold provides insight into the dynamics of the underlying system.  I present a manifold learning framework for biomedical time series analysis, and I provide a guarantee on the dynamical information which can be recovered using this framework.   We will discuss several applications in the analysis of electrocardiograms and blood pressure monitoring.  Time-permitting, we will examine further approaches to biomedical time series analysis which fuse traditional signal processing with modern mathematics.