Uploaded October 2026 | Updated November 2026, 3 hours ago
Speaker: Kenji Doya, Professor, Neural Computation Unit
Title: Reinforcement learning in the brain and robots
Abstract:
Reinforcement learning is a theoretical framework for an active agent, either human, animal, robot, or software, to learn actions guided by reward signal evaluating its performance. Since the start or our Neural Computation Unit in 2004, revealing the brain’s mechanisms for reinforcement learning and developing more flexible reinforcement learning mechanisms for robots have been the major dual goals.
In this lecture, I will present our works till today about how the basal ganglia evaluate the goodness of actions, how the network linking the cortex, the basal ganglia and the cerebellum can realize action planning by mental simulation, and how the serotonergic system balance short- and long-term evaluation. I will also introduce our ongoing works on how prediction and control can be realized in the cortical circuits, and how rewards can be evolved for survival, reproduction, and curiosity.
The Provost Lecture Series will provide an opportunity to celebrate milestones in the careers of OIST faculty members. It will recognize newly promoted professors and those who have received prestigious awards. It will highlight faculty achievements and create a space for retiring professors to talk about their research accomplishments and lessons learned over the course of their career. It is a way to celebrate OIST faculty, introduce them to the broader OIST community, acknowledge their research successes, share about their teaching and mentorship of students, and honor their service to the university.
Speaker: Kenji Doya, Professor, Neural Computation Unit
Title: Reinforcement learning in the brain and robots
Abstract:
Reinforcement learning is a theoretical framework for an active agent, either human, animal, robot, or software, to learn actions guided by reward signal evaluating its performance. Since the start or our Neural Computation Unit in 2004, revealing the brain’s mechanisms for reinforcement learning and developing more flexible reinforcement learning mechanisms for robots have been the major dual goals.
In this lecture, I will present our works till today about how the basal ganglia evaluate the goodness of actions, how the network linking the cortex, the basal ganglia and the cerebellum can realize action planning by mental simulation, and how the serotonergic system balance short- and long-term evaluation. I will also introduce our ongoing works on how prediction and control can be realized in the cortical circuits, and how rewards can be evolved for survival, reproduction, and curiosity.
The Provost Lecture Series will provide an opportunity to celebrate milestones in the careers of OIST faculty members. It will recognize newly promoted professors and those who have received prestigious awards. It will highlight faculty achievements and create a space for retiring professors to talk about their research accomplishments and lessons learned over the course of their career. It is a way to celebrate OIST faculty, introduce them to the broader OIST community, acknowledge their research successes, share about their teaching and mentorship of students, and honor their service to the university.










