CAREER: State Space Models for Neural and Behavioral Data

About This Grant

Large-scale brain recording technologies now enable the activities of thousands of neurons to be monitored simultaneously in freely moving animals. In parallel, advances in computer vision enable fast and accurate tracking of animal behavior, which offers rich opportunities for evaluation of the nervous system's output. Capitalizing on these new technologies, this project focuses on critical statistical challenges that must be overcome to tackle a central challenge of neuroscience: linking brain activity to behavioral output. Neural and behavioral data are noisy, high-dimensional time series with complex dynamics. This project will develop novel computational and statistical methods to analyze these data and offer new insight into how the brain produces natural behavior. Specifically, this project will develop state space models (SSMs) — statistical models for high-dimensional time series data — to link brain activity and behavior. The first objective is to develop SSMs for neural recordings. These methods will allow scientists to model how brain activity changes across contexts, how it is influenced by chemical signals like neuromodulators and neuropeptides, and how it can be precisely modulated through real-time, closed-loop experiments. The second objective is to develop similar methods for the study of behavioral video data, allowing scientists to quantitatively capture how behavior is shaped by modulatory inputs, how it changes throughout aging, and how behavior is structured in pursuit of latent goals. Finally, this project will leverage state-of-the-art methods in artificial intelligence called deep SSMs, tailoring them for use in brain-computer interfaces, and developing a theory of how deep SSMs work. Altogether, this project will contribute novel methods that will provide new insight into how the brain produces rich, natural behavior. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.