Shinichiro Tsutsumi

The paper shows that climbing fiber inputs in the cerebellum generate reward-prediction errors that guide modular reinforcement learning in mice, integrating with supervised learning mechanisms.

The paper analyzes two-photon imaging data from mice during reinforcement learning to show cerebellar climbing fiber responses are organized into four functional components that reduce learning space dimensions.

The paper shows that cerebellar climbing fibers encode multiple behavioral variables and are organized into functional modules in mice.

This paper introduces HA_time, a hyperacuity algorithm for precise spike timing estimation from calcium imaging, achieving up to 20-fold improvement in spike time precision.