Neural representations of learning and behavior

A fundamental question in neuroscience is to understand how coordinated neural activity and structured circuitry in the brain are responsible for generating behavior. Decades of experimental and theoretical work have demonstrated that correlated neural activity has a strong impact on population coding by reshaping neural representations of external stimuli. More recently, large-scale recordings have provided insight as to the dynamic mechanisms by which neural populations perform computations, but the link to their circuitry remains unclear. A promising region for understanding the relationship between neural circuitry, population activity, and behavior is the cerebellum, whose evolutionarily-conserved, crystalline circuitry is the basis of a critical role in motor control and learning guided by sensory errors. In this thesis, I provide an overview of my contributions in neuroscience which straddle multiple quantitative approaches: from abstract theoretical modeling of neural networks, to analysis of cerebellar recordings in close collaboration with experimental colleagues, and finally the development of new data-driven methods to identify behaviorally-relevant information in high-dimensional neural recordings.