Computation & Interpretation

In the last few years, many new computational methods have been created to make sense of the large amounts of neural population data collected using new high-density recording techniques. However, many of these methods suffer from criticial limitations: 1) being under-constrained neural data alone, 2) not disambiguating the roles of different brain regions, and 3) not generating the control signals necessary for movement. These methods struggle to predict inputs from other brain areas or sensory feedback, as too many possible solutions exist. To constrain model solutions and predict the interaction of brain areas, we are developing frameworks for jointly modeling neural activity and skilled closed-loop motor control. These new methods allow us to describe the contributions of different brain regions to closed-loop motor control and characterize the information communicated between regions.