Abstract

Some brain disorders are hypothesized to have a dynamical origin; in particular, it has been hypothesized that some symptoms of Parkinson's disease are due to pathologically synchronized neural activity in the basal ganglia region of the brain. We construct artificial neural networks to first demonstrate the capacity to make meaningful predictions of bursting events in patient Parkinsonian data, then use the same framework to generate an adaptive control strategy applied to computational modeling. Our model shows meaningful success in predicting the future brain state in Parkinsonian patients compared to alternative approaches. With this baseline established, we demonstrate that the network can quickly learn precise control and maintain high fidelity to the control objective in a simulated environment, even if the underlying model's parameters vary.

Bio

Jeff Moehlis received a Ph.D. in Physics from UC Berkeley in 2000, and was a Postdoctoral Researcher in Applied and Computational Mathematics at Princeton University from 2000-2003. He joined the Department of Mechanical Engineering at UC Santa Barbara in 2003, and is currently department Chair. He was also recently the Chair of the Program in Dynamical Neuroscience at UC Santa Barbara. He has been a recipient of a Sloan Research Fellowship in Mathematics and a National Science Foundation CAREER Award, and was Program Director of the SIAM Activity Group in Dynamical Systems from 2008-2009.  He has supervised 11 students to completion of a PhD degree, and 7 additional students who received an MS degree. Jeff's current research includes applications of dynamical systems, control, and computational techniques to neuroscience and system identification. He has published approximately 120 journal/conference proceedings articles on these and other topics including cardiac dynamics, collective behavior, shear flow turbulence, microelectromechanical systems, energy harvesting, and dynamical systems with symmetry.

Abstract

Some brain disorders are hypothesized to have a dynamical origin; in particular, it has been hypothesized that some symptoms of Parkinson's disease are due to pathologically synchronized neural activity in the basal ganglia region of the brain. We construct artificial neural networks to first demonstrate the capacity to make meaningful predictions of bursting events in patient Parkinsonian data, then use the same framework to generate an adaptive control strategy applied to computational modeling. Our model shows meaningful success in predicting the future brain state in Parkinsonian patients compared to alternative approaches. With this baseline established, we demonstrate that the network can quickly learn precise control and maintain high fidelity to the control objective in a simulated environment, even if the underlying model's parameters vary.

Bio

Jeff Moehlis received a Ph.D. in Physics from UC Berkeley in 2000, and was a Postdoctoral Researcher in Applied and Computational Mathematics at Princeton University from 2000-2003. He joined the Department of Mechanical Engineering at UC Santa Barbara in 2003, and is currently department Chair. He was also recently the Chair of the Program in Dynamical Neuroscience at UC Santa Barbara. He has been a recipient of a Sloan Research Fellowship in Mathematics and a National Science Foundation CAREER Award, and was Program Director of the SIAM Activity Group in Dynamical Systems from 2008-2009.  He has supervised 11 students to completion of a PhD degree, and 7 additional students who received an MS degree. Jeff's current research includes applications of dynamical systems, control, and computational techniques to neuroscience and system identification. He has published approximately 120 journal/conference proceedings articles on these and other topics including cardiac dynamics, collective behavior, shear flow turbulence, microelectromechanical systems, energy harvesting, and dynamical systems with symmetry.