Our laboratory focuses on understanding the neural mechanisms for decision making, reinforcement learning, and number sense in healthy and disease states. We investigate these mechanisms through large-scale recordings of neural activity and synaptic plasticity dynamics from mice engaged in sophisticated head-fixed behavior tasks. We employ data science approaches, including the use of AI, to analyze and simulate animal behaviors and neural network dynamics.
Projects
Reinforcement Learning (RL)
RL is a trial-and-error learning that involves interactions with an environment. Through the interactions, animals learn optimal behaviors by reinforcing actions that lead to rewarding events and avoiding actions that lead to undesired outcomes. RL has been also very powerful in training artificial intelligence (AI), enabling AI to outperform humans in complex tasks. Understanding of the neural mechanisms of RL facilitates not only the development of treatment for learning and cognitive impairments but also development of novel RL approach in the field of AI. However, the underlying neural mechanisms are complex because the processes involve many brain areas. We investigate how different brain areas interact to process the computations that are necessary for the RL and its associated value-based decision making, and how these processes may be impaired in neurological disorders.
Number Sense
Number is an important concept in our daily lives as humans. However, the neural mechanisms for number sense have been poorly understood. We will investigate how number information is processed in the brain.
Approach
Virtual Reality (VR)
Flexibility in VR allows designing sophisticated head-fixed behavior tasks for mice. We develop VR behavior tasks to investigate the neural mechanisms underlying cognitive flexibility and learning.
Large-scale 2-photon imaging
We use a large field-of-view (FOV) 2-photon microscope to image neural activity and plasticity dynamics that underly cognitive functions and learning. We can record neural dynamics across distant brain areas and analyze their inter-areal communications.
Optogenetics
To examine the functions of neural activity dynamics in a brain area, we manipulate the activity dynamics using optogenetics. By expressing light-sensitive opsins in neurons, we can manipulate their neural activity with light illumination and reveal the functions of each brain area in mouse behaviors.
Data Science, Artificial Intelligence (AI)
Our experimental approaches generate big data. How do we interpret the complex animal behaviors and neural population dynamics? We employ data science approaches, including AI, to analyze and interpret the big data from experiments. We also train AI models in simulated VR tasks to understand the similarities and differences between animal brains and AI models.
