Friday,
April 29, 12:00 p.m. PSA 113
Robert Thompson
School of Mathematical and Statistical Sciences
Reverse-Engineering Dynamical Systems From Time Series Data Using
Compressed Sensing Techniques
Abstract
The theory of compressed sensing focuses on reconstructing signals from
a relatively small number of samples, when certain signal sparsity
constraints are satisfied. We will discuss some empirical explorations of an
innovative application of this theory, proposed by Wang, Yang, Lai, Kovanis,
and Grebogi: the ability to recover the system of equations governing a
dynamical system from a relatively small number of time-series data points. In
the context of some canonical examples we will look at when it works, when it
does not, and some initial thoughts on how compressed sensing theory might be
able to illuminate what is happening.
