CRCNS: US-Japan Research Proposal: The Computational Principles of a Neural Face Processing System

在我们对计算原理和神经机制的理解上存在着根本性的差距，神经电路通过这些原理和机制来代表复杂的物体，比如脸。这一概念空白构成了一个重要的问题，因为在填补它之前，我们将无法理解人脸识别和人脸失明的原因。长期目标是了解人脸识别的计算原理和神经机制，并基于这些原理创建计算机人脸识别系统。这项建议的总体目标是确定大脑脸部处理网络中局部和全局脸部特征编码的计算原则。中心假设是，面孔选择区域的高水平特征调整可以理解为刺激空间的统计特性和处理这些刺激的回路的一般组织特征的结果。这一提议的基本原理是，拟议研究的完成将提供对神经回路如何产生复杂视觉形状的有意义的表征的理解，从而对视觉理论施加关键限制。这一假设将通过追求三个具体目标来检验，这三个目标将决定1)面部特征调整的神经机制，2)分类面孔选择的神经机制，3)特征调整转换的神经机制。联合计算和实验方法将把定位面部区域的功能磁共振成像与针对这些区域的电生理记录和模型系统的计算分析结合起来。该方法通过理论原理和实验验证的紧密耦合以及通过开发新的理论和实验方法而具有创新性。这项研究意义重大，因为它将揭开神经回路功能的原理，这些原理对于理解视觉对象识别和多节点网络具有普遍意义。由于这一成果是理解社会知觉回路机制的进步，它将识别面部处理系统的脆弱性，直接与理解面孔失明、面容失认症以及自闭症谱系障碍、脆性X综合征和威廉姆斯综合征等综合征的社会知觉改变有关。

There is a fundamental gap in our understanding of the computational principles and neural mechanisms by which neural circuits represent complex objects like faces. This conceptual gap constitutes an important problem because, until it is filled, we will not be able to understand face recognition and the reasons for face blindness. The long-term goal is to understand the computational principles and neural mechanisms of face recognition and create a computer face-recognition system based on these principles. The overall objective of this proposal is the determination of the computational principles of local and global face feature coding in the brain's face-processing network. The central hypothesis is that high-level feature tuning in face-selective areas can be understood as the result of the statistical properties of stimulus space and general organizational features of the circuits that process these stimuli. The rationale for this proposal is that completion of the proposed research will provide an understanding, of how neural circuits generate a meaningful representation of complex visual shape, imposing critical constraints on theories of vision. The hypothesis will be tested by pursuing three specific aims, which will determine 1) neural mechanism for facial feature tuning, 2) neural mechanism for categorical face selectivity, and 3) neural mechanisms for transformations of feature tuning. The joint computational and experimental approach will integrate functional magnetic resonance imaging to localize face areas with electrophysiological recordings targeted to these regions and computational analyses of model systems. The approach is innovative through the tight coupling of theoretical principles and experimental validation and by developing novel theoretical and experimental methodologies. The proposed research is significant, because it will unravel principles of neural circuit function that are of general relevance for understanding visual object recognition and multi-node networks. Because the outcome is an advance in understanding circuit mechanisms of social perception, it will identify vulnerabilities of the face-processing system directly relevant to the understanding of face blindness, prosopagnosia, and of altered social perception in syndromes spanning autism spectrum disorders, fragile X, and Williams syndrome.