入力および出力に関する脳情報処理のベイズ的解釈と数理モデル化

In this paper, we explain the Bayesian inference framework for understanding the computational principles underlying the brains functions. We show that this framework possibly gives a unified theory for two types of information processing: input-related computation and output-related one. We also show that the two types of information processing can be mathematically formulated as probabilistic inference problems on a dynamic Bayesian network. This model provides theoretical foundation such as Bayes theorem on output-related computation, which may be useful for realization of artificial brain-like information processing systems.