An integrative study on adaptation of human perception based on Bayesian models

人类的大脑面临着外部世界和神经系统的各种不确定性。众所周知，我们可以利用直接可观测量（称为似然）和观测之前已知的信息来处理这种不确定性。我们如何学习这些信息？在本研究中，我们从理论上和实验上探讨了这两个因素的学习是如何相互关联的，特别是可能性是如何学习的，这是迄今为止很少有研究关注的问题。我们展示了学习这两个因素的不同之处，并展示了学习可能性是非常有效和灵活的。

The human brain faces many kinds of uncertainty in the outer world and the nervous systems. It has been known that we can deal with such uncertainty using information from directly observable quantities (called likelihood) and information that is known prior to the observation. How can we learn those information? In this study, we theoretically and experimentally investigate how the learning of those two factors relate to each other, and especially how the likelihood is learned, which is the problem that few studies has put focus on so far. We showed how the learning of the two factors differed, and showed that the likelihood is learned very effectively and flexibly.