Title: Using Archetypes to Analyze and Control Complex Spatio-Temporal Dynamics
Speaker: Professor Emily Stone
Speaker Info: Utah State University
Brief Description:
Special Note:
Abstract:
Archetypal analysis is a new statistical technique for extracting "archetypal patterns" from data sets. It is related to principal component analysis (a.k.a. POD, KL), though its objectives are different: it finds data points on the convex hull of the data set, rather than eigendirections in the data space.Date: Friday, November 7, 1997A variation of archetypal analysis developed for tracking traveling structures (simple examples being solitons or fronts) has been developed, called "moving archetypes". We have used moving archetypes to reduce a spatially/temporally complex regime from the Kuramoto-Shivashinsky equation to a set of time series, in a manner similar to POD. The time series are then used to enact Ott-Grebogi-Yorke type control on the PDE, returning the solution to a "laminar" regime- a modulated traveling wave. The details of the techniques and the success and limitations of the method will be discussed.