The Wu Tsai Neurosciences Institute and Stanford Data Science are pleased to announce a special seminar series focused on the intersection of data and brain science.
Dynamical mechanisms underlying robust and flexible computation in neural populations
How do neural circuits robustly and flexibly perform computations that enable sensory perception, decision making, and motor control? It is thought that such computations are implemented through the dynamical evolution of distributed activity in recurrent circuits. Thus, a key goal is to characterize from population recordings how this computational activity evolves, and how the dynamical properties of the circuit give rise to this evolution and thus the computation. In this talk, I’ll present work that addresses these questions in the primate motor cortex.
I’ll present a novel analytic approach that relates measured neural activity to theoretically tractable, dynamical models of excitatory and inhibitory neurons. I’ll show applications to the analysis of motor cortical population responses to optogenetic and electrical microstimulation perturbations during reaching behavior. These analyses reveal that motor cortical activity during reaching is shaped by a self-contained, low-dimensional dynamical system. The subspace containing task-relevant dynamics is oriented so as to be robust to strong non-normal amplification within cortical circuits. Stimulation in the motor cortex perturbs reach-kinematics only to the extent that it alters neural states within this subspace, suggesting that the task dynamics space exhibits a privileged causal relationship with behavior.
These results resolve long-standing questions about the dynamical structure of cortical activity associated with movement, provide links between low-dimensional structure in neural population activity and mechanistic interpretations thereof, and illuminate the dynamical perturbation experiments needed to understand how neural circuits generate complex behavior.
Lea Duncker is a postdoctoral researcher at Stanford University, where she has worked in the Shenoy and Linderman Labs. She has a bachelor of science in natural sciences and a master of science in computational statistics and machine learning, both from University College London. Duncker obtained her Ph.D. at the Gatsby Computational Neuroscience Unit under the supervision of Maneesh Sahani. Her research focuses on characterizing and interpreting population-level structure in neural data, with a focus on dynamical systems and motor control.