摘要
研究人■神经网络的信息处理机制是模拟人脑功能要解决的一个重要问题,在信息几何框架(?),该机制可以通过研究信息系统的几何结构来讨论.本文提出了一种简化的模拟人类思维层次的层次化神经系统模型.并利用基(?)信息几何的神经场学习理论解释了不同层次的神经系统通过(?)馈和反馈连接进行动态交互作用的逼近学习机制,进而从整体宏观的角度对人脑学习机理以及概念形成给出了(?)个数学(?)
Studying the information processing mechanism of the artificial neural networks is an important problem to simulate the function of human brain. This mechanism can be discussed by studying the geometric structure of information system in the framework of information geometry. This paper presents a simplified hierarchical neural system model to simulate human s thinking hierarchy and explains the feedforward and feedback dynamic interactions between the hierarchies using the neural field learning theory based on information geometry. This paper gives a mathematical presentation of brain s learning mechanism and the formation of concept.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第3期332-336,共5页
Pattern Recognition and Artificial Intelligence
基金
教育部博士点基金(No.20020004020)
关键词
信息几何
神经场
层次化
学习机制
Information Geometry
Neural Field
Hierarchy
Learning Mechanism