Naoshige Uchida, Ph.D. (Harvard)

Professor of Molecular and Cellular Biology
Center for Brain Science
Harvard University 

“Diversity of dopamine neurons: multi-agent reinforcement learning”

meeting ID: 914 7618 4522


Dopamine regulates multiple brain functions including learning, motivation and movement. Furthermore, the striatum, a major target of dopamine neurons, is parceled into multiple subregions that are associated with different types of behavior, such as Pavlovian, goal-directed, and habitual behaviors. An important question in the field is how dopamine regulates these diverse functions. It has been thought that midbrain dopamine neurons broadcast reward prediction error signals to drive reinforcement learning. However, recent studies have found more diverse dopamine signals than originally thought. How can we reconcile these results? In this talk, I will discuss our recent studies characterizing diverse dopamine signals, and how these findings can be understood in a coherent theoretical framework.


Speaker Bio

Naoshige Uchida received his Ph.D. on his study on the molecular mechanism of synaptic adhesions in Masatoshi Takeichi‘s laboratory at Kyoto University, Japan. He studied olfactory coding in Kensaku Mori‘s laboratory at the Brain Science Institute, RIKEN, Japan. He then joined Zachary F. Mainen‘s laboratory at Cold Spring Harbor Laboratory, New York, USA, where he developed psychophysical olfactory decision tasks in rodents. He started his laboratory at Harvard University in 2006.

His long-term research goal is to understand the neural mechanisms underlying perception and decision-making. To achieve this goal, his laboratory integrates various techniques including neuronal recording in behaving animals, psychophysical behavior experiments, optogenetics, and virus-mediated neural circuit tracing in rodents. He has made a number of findings in his studies on the midbrain dopamine system which combined computational modeling using reinforcement learning theories and electrophysiology in behaving mice. To conduct these studies, he developed a method to optogenetically identify genetically-defined neuronal populations during recording as well as a method to identify monosynaptic inputs to genetically-defined neuronal population anatomically. He has also studied the effect of state uncertainty and belief states in reinforcement learning, as well as a novel algorithm developed in artificial intelligence, called distributional reinforcement learning.


Feb 22 2022


2:40 PM - 4:00 PM


Location: 19 West 4th Street, Room 517, New York, NY


Categories: Neuroeconomics Colloquium