ECE Departmental Seminar
Learning tree-structured neural dynamics from single trial population recording
Prof. Memming Park
Stony Brook University, Neurobiology and Behavior
Friday, 10/26/18, 11:00am
Light Engineering 250
Abstract: To understand the complex nonlinear dynamics of neural circuits, we fit a structured state-space model to noisy high-dimensional neural time series. We use Bayesian inference methods and an initialization strategy for both the posterior state trajectory and the posterior dynamics. In this work, we consider a multi-scale hierarchical generative model for the state-space dynamics. Each node of the latent tree captures locally linear dynamics. The expressive power and interpretability of the inferred dynamics are evaluated for neural computation systems.
Bio: Il Memming Park is an Assistant Professor in Neurobiology and Behavior at Stony Brook University. He is a computational neuroscientist trained in signal processing, statistical modeling, information theory, and machine learning. He specializes in analyzing neural signals. He obtained his M.S. in electrical engineering, Ph.D. in biomedical engineering from University of Florida, and B.S. in computer science from KAIST.