Neural and cognitive mechanisms of predictive coding and their interactions for perception and action

在日常生活中，我们很少会对自己的能力感到惊讶，比如能拿一杯咖啡，能认出朋友的脸，或者能在公园里慢跑。但这些任务具有巨大的挑战性：我们必须处理大量模糊和不完整的信息进入我们的眼睛和其他感官。那么，为什么我们不总是洒咖啡，忽视我们的朋友或撞到树上？越来越多的证据表明，为了应对这些挑战，大脑的行为就像一个侦探，做出聪明的推断。大脑不仅使用可用的感官“证据”，而且还使用关于世界的先验知识来预测世界应该是什么样的。其基本原理被称为预测编码，最近的计算机模型提出，视觉皮层可能在所谓的预测模块的层次结构中实现该原理。然而，目前还不清楚，同一个等级系统是否可以解释任何形式的预测机制，这些预测机制是基于对世界的无意识、内隐预期以及有意识或外显预期，或者是否适用于不同的机制。同样不清楚的是，视觉过程是如何与视觉功能联系在一起的。因此，我建议研究视觉感知与几种形式的期望，以及在手和眼运动的背景下的视觉。领导一个研究生和本科生的研究团队，我将使用“预期特征感知范式”测试健康人类参与者的物体感知，同时记录来自大脑的电生理信号或用磁脉冲刺激大脑功能。* 主题1将测试明确期望的影响（即，注意颜色或形状）和隐含的期望（例如，由于刺激的重复呈现而引起的瞬时可塑性）在如预测编码模型所预测的电生理脑信号的特定时间相互作用，或者是否需要更复杂的模型。* 主题2将深入研究注意力在注意力能力和刺激相似性方面的预测效果。* 主题3将研究视觉引导抓握运动期间的预测编码原则。* 主题4将研究眼动过程中的预测编码。* 总而言之，这项研究将寻求一种新的、全面的方法来研究视力。这项研究有望让我们更全面地了解大脑中不同形式的预测编码。这样的理解将是迈向视觉大脑作为嵌入复杂世界的动态系统的统一理论的重要一步，对增强现实等技术应用具有重要意义。

In daily life we seldom marvel at abilities such as grabbing a cup of coffee, recognizing a friend's face, or jogging through a park. But these tasks are enormously challenging: we have to deal with overwhelming amounts of ambiguous and incomplete information entering our eyes and other senses. Why then don't we always spill coffee, ignore our friends or crash into trees? There is growing evidence that to deal with these challenges the brain behaves much like a detective who makes smart inferences. The brain uses not only available sensory "evidence," but also prior knowledge about the world to predict what the world should be like. The underlying principle is called predictive coding and recent computer models have proposed that the visual cortex may implement the principle in a hierarchy of so-called prediction modules. It is unclear, however, whether the same hierarchical system can explain any form of prediction mechanisms based on unconscious, implicit expectations as well as conscious or explicit expectations about the world, or whether different mechanisms apply. It is also unclear how visual process-es tie in with visuomotor functions. Therefore, I propose to study visual perception together with several forms of expectations, as well as vision in the context of hand and eye movements. Leading a research team of graduate and undergraduate students I will test object perception in healthy human participants using an `expected feature perception paradigm' while recording electrophysiological signals from their brain or while stimulating brain functions with magnetic pulses. ***Theme 1 will test whether effects of explicit expectations (i.e., attention to colours or shapes) and implicit expectations (e.g., transient plasticity due to repeated presentation of stimuli) interact at specific times of electrophysiological brain signals as predicted by predictive coding models or whether more complex models are required. ***Theme 2 will investigate in depth predictive effects of attention in terms of attentional capacity and stimulus similarities. ***Theme 3 will examine predictive coding principles during visually guided grasp movements. ***Theme 4 will investigate predictive coding during eye movements. ***In sum, this research will pursue a new, comprehensive approach to investigating vision. This research can be expected to give us a more complete understanding of the different forms of predictive coding in the brain. Such understanding would constitute a significant step towards a unifying theory of the visual brain as a dynamic system embedded in a complex world with important benefits for technical applications such as augmented reality.**