A03 Understanding the sensory and motor cortical circuits based on the duality of inference and control
The cerebral cortex has a standard six-layer structure called ‘the canonical circuit’ spanning from the sensory cortices to the motor and prefrontal cortices. However, it has not been answered, or even questioned, how a common circuit can perform different functions like sensory cognition and motor control. On the other hand, a recent theoretical framework of ‘control as inference’ has revealed that Bayesian inference for perception and reinforcement learning for control can be realized by common computational procedures, which has contributed to development of novel algorithms and their applications.
Based on the duality of Bayesian inference and reinforcement learning, this research considers hypotheses about how cortical circuits can implement algorithms like message passing and variational free energy approximation and test them by computer simulations and mouse neural recordings.