【摘要】 通过对人的相对低层次认知行为的模仿可以改善计算机感知外部世界的能力 .知觉作为首要的认知行为 ,它所需要的知识不便于采用传统的符号化知识表示方法来表示 .该文以视知觉研究为基础 ,通过模拟分布于视皮层中的特异性功能柱型结构 ,将图形模式分解为简单特征的组合 ,直接表示在神经网络上 ,众多功能单元的同时响应就构成某一实体的视觉表象 .最后还讨论了这种直接知识表示方法在知识与概念化问题、知识的来源问题、学习问题等几个最核心的人工智能问题下的意义 .更多还原

【Abstract】 The ability of computers to perceive outside world is very poor, that could be improved by simulating human beings’ low level cognitive function. Perception is the most primary skill of cognition, the knowledge what it needed to make order out of a chaos of physical stimulation——a senseless, meaningless milieu, to achieve meaningful knowledge of the physical universe, is not appropriate to be represented by formal symbols. On the base of visual perception research, a new method of naive mental image represented distributedly among neural network, which is a hierarchical model of visual center cortex, is discussed in this paper. There are no cognition units representing the concept of objects but the concept of objects is only represented by the activities distributed over various regions of units. The significance of this direct knowledge representation for the most kernel problems, such as pre eminence of knowledge and conceptualization,disembodiment,learning separation,kinematics of cognition, are also considered at the end of this paper.更多还原