Artificial neural network for understanding neural basis of mouse decision making by Akihiro Funamizu (Lecturer, Institute for Quantitative Biosciences (IQB), University of Tokyo)
Our brain has multiple strategies and changes the strategy in a context-dependent manner. For example, model-free strategy decides choices based on direct experiences of choices and rewards, while inference-based strategy decides choices by estimating the hidden context from observed sensory inputs. A field of systems neuroscience often hypothesizes that the brain has parallel neural circuits to drive different strategies in cortico-basal ganglia circuit. In contrast, a field of artificial intelligence (AI) shows that one recurrent neural network can model choice behaviors in multiple strategies, providing a hypothesis that one neural circuit, instead of parallel circuits, can drive various behavior. Our research investigates the neural basis of how mice use different strategies for making choices by combining electrophysiology and behavioral modeling. Today, I am going to talk about our recent challenges of analyzing mouse choice behavior with artificial neural network.