Jack Lindsey

Hi, I’m Jack, a second-year Ph.D. student in computational neuroscience at Columbia. I did my undergrad at Stanford University in math and computer science. I’m interested in understanding how neural representations support complex behaviors and how these capabilities are learned. Some projects I’m working on include: (1) modeling how learned associations and skills are consolidated into long-term memory; (2) understanding the role that heterogeneous dopamine activity plays in reinforcement learning; and (3) exploring how recurrent dynamics in sensory representations affect associative learning. I am applying these modeling ideas to particular learning systems in insects (e.g. the mushroom body) and mammals (e.g. striatum). I also like thinking about how modern machine learning (ML) approaches inform our understanding of these systems, and conversely how neural systems may suggest improvements to ML algorithms. Outside of research, I enjoy tennis, playing guitar, reading, eating cookies and making puns.

Principal Investigators: Ashok Litwin-Kumar & Larry Abbott

Fellow: Jessica Lee

Project
Recent advances in connectomics have enabled the construction of a nearly complete map of neuronal connectivity in the fly brain, containing over twenty million synapses. The unprecedented scale and detail of this data presents a computational challenge: How can we make sense of this neural circuitry and understand its function? In this project, the fellow will work on developing and applying algorithms to identify functionally important pathways and hubs in the connectome data. We are particularly interested in the mushroom body, a key site of learning and memory in the fly brain. Preliminary analysis indicates that dopaminergic neurons (DANs) in the mushroom body, which modulate synaptic plasticity, receive highly diverse inputs from across the brain. Untangling the pathways that drive DAN activity will allow us to make predictions about how flies learn, and may reveal parallels (or inspire extensions) to modern reinforcement learning algorithms.