Computational Neuroscience, Brandenburg University of Technology Cottbus-Senftenberg, Cottbus, Germany.
Prog Brain Res. 2019;249:3-18. doi: 10.1016/bs.pbr.2019.04.025. Epub 2019 Jun 6.
Sequential Bayesian updating has been proposed as model for explaining various systematic biases in human perception, such as the central tendency, range effects, and serial dependence. The present chapter introduces to the principal ideas behind Bayesian updating for the random-change model introduced previously and shows how to implement sequential updating using the exact method via probability distributions, the Kalman filter for Gaussian distributions, and a particle filter for approximate sequential updating. Finally, it is demonstrated how to couple perception to action by selecting an appropriate action based on the posterior distribution that results from sequential updating.
序贯贝叶斯更新已被提议作为解释人类感知中各种系统偏差的模型，例如中心趋势、范围效应和序列依赖。本章介绍了先前介绍的随机变化模型的贝叶斯更新的主要思想，并展示了如何使用概率分布的精确方法、高斯分布的卡尔曼滤波器和用于近似序贯更新的粒子滤波器来实现序贯更新。最后，通过根据序贯更新产生的后验分布选择适当的动作，演示了如何通过将感知与动作耦合来实现这一点。
