Using Reservoir Computing to Detect and Suppress Disturbances to Dynamical Systems

Abstract: Identifying and suppressing unknown disturbances to dynamical systems is a problem with applications in many different fields.  In this talk, I will present a model-free method to identify and suppress, in real-time, an unknown disturbance to an unknown system based only on previous observations of the system under the influence of a known forcing function.  The method is based on reservoir computing, a machine learning architecture especially suited for dynamical systems.  Under very mild restrictions on the training function, the method is able to robustly identify and suppress a large class of unknown disturbances.  I will illustrate the scheme with an example where chaotic and stochastic disturbances to the Lorenz system are identified and suppressed.  In addition, I will show that the method can be applied to network-coupled dynamical systems, and used to identify which nodes in the network are perturbed.  Finally, I will discuss how the method can be applied to large networks by using a parallelization scheme. 

Speaker’s Bio: Juan G. Restrepo is an Associate Professor of Applied Mathematics at the University of Colorado at Boulder.  After studying physics as an undergraduate in Colombia, he completed his Ph.D. in Applied Mathematics at the University of Maryland in 2005.  After that, he spent two years as a postdoctoral student at Northeastern University and has been at the University of Colorado at Boulder since 2008.  Prof. Restrepo studies synchronization, nonlinear dynamics and chaos, dynamics on and the structure of complex networks, and the spread of information and disease on networks.