I'm currently working on studies related to Bayesian neural computation. This field dates back at least to the 19th century where von Helmholtz proposed an idea of "unconscious inference", postulating that perception is a process of unconsciously inferring something that we can't see from something that we can see. A modern computational neuroscientific counterpart of this conceptual claim would be the Bayesian brain hypothesis, which considers the brain somehow conducts Bayesian inference, a mathematical (statistical or probabilistic) formulation of estimation of hidden cause of current observations incorporating prior knowledge and past experience.
The Bayesian brain hypothesis has been quite influential in this field of computational neuroscience. The somehow has been broken down into more specific frameworks that allow systems to exactly or approximately conduct Bayesian computation, and the conditions that derives biologically appealing physical implementations of such computation has been explored.
My recent studies are pursuing further biological plausibility of Bayesian neural computations. An preliminary study (Kataoka & Doya, 2026, arXiv) has introduced broader statistical properties to predictive coding, one of the neural implementations of Bayesian inference, leading to nonlinear and heterogeneous electrophysiological characteristics that has been observed experimentally.
The tissue organisation covering the mammalian brain, called neocortex, is believed to play significant functional roles. Visual and auditory cortices process corresponding sensory information, interacting with "higher-order" areas such as prefrontal and other association cortices, and of course with the subcortical regions beneath the cortical tissue such as basal ganglia, thalamus, or hippocampus.
My recent studies are interested in the fact that different cortical areas share surprisingly similar microcircuit structure over the neocortex. The microscopic circuit structures are known not to differ that much between lower and higher stages of visual cortices, between visual and auditory cortices, and between sensory (including visual and auditory areas) and motor-controlling cortices. I'm exploring the computational principles, especially the ones regulating learning, that explains the mechanisms and functions of the sensory and motor cortices in a unified manner; although we know that the functions of these areas are involved are fundamentally different, their principal mechanisms should be shared given the structural similarity of microcircuitry of these areas!
I'm funded by a 3-year grant for exploration of such shared mechanisms (JSPS KAKENHI 26KJ0370).