Anne Draelos
Department of Biostatistics and Bioinformatics
Duke University

"Scalable online modeling and perturbations for adaptive neuroscience experiments"

Nov 16, 2021 Schedule:

08:15 to 09:15 AM Eastern Time (US and Canada)

Meeting # 989641629322


An important class of models in systems neuroscience are those that leverage new recording technologies and statistical analyses to describe the coordinated activity of larger populations of neurons, even entire brains. Combined with new, precise stimulation technologies, we could begin to dissect large-scale circuits in vivo, constructing models that causally relate neural activity to behavior. Direct testing of hypothesized brain-behavior links, however, requires statistically efficient methods for both estimating and intervening on population-level neural dynamics in real time. To build neural models online, we describe a new method that combines fast, stable dimensionality reduction with a soft tiling of the resulting neural manifold, allowing dynamics to be approximated as a probability flow between tiles. This method can be fit efficiently, scales to large populations, and outperforms existing methods when dynamics are noise-dominated or feature multi-modal transition probabilities. Next, we present an algorithm for selecting optimal circuit interventions to create maps of causal influence within large populations of neurons. This method uses online convex optimization and adaptive stimulation selection to quickly infer the binary network connectivity, rendering the inference of networks of tens of thousands of neurons in vivo feasible in a single experiment. Our efficient and scalable methods for online brain state modeling and stimulation are fast enough to make real-time, causal tests of neural function feasible at scale.