摘要
针对具有多层次性和复杂性的认知问题提出一个动态可增殖的多层次自组织认知系统,每个层次具有形式上一致的知识表示方法,各层的自组关联、自组聚合、归约和样本表达四个知识处理模型是实现系统自组织层次增殖的核心模型。提出信息粒和容器的概念,信息粒演进流程模拟认知过程的静态归约,容器演进流程对应于认知的动态模拟,这两个流程在每个系统层次上对每个输入样本完成一个完整的模拟认知与归约表达。以自组图演算为理论模型,给出了每个层次内以及相邻层次之间的信息处理与传递的详细设计规范。
For the hierarchical and complex recognition problems,a generality model for dynamical grow recognition system with hierarchical self-organization is proposed.There has a formal unity knowledge representation in every layer. There are four knowledge process models for each layer,the first model is self-organizing connection model,the second is self-organizing aggregation model,the third is reduction model and the last is sample representation model.These four models are the core algorithms for the dynamical grow recognition system with hierarchical self-organization.The concepts of information grains and containers are given.Information grains process circuit simulate the static reduction of recognition,and containers process circuit simulate the dynamic procedure of recognition.Through these two circuits,a complete recognition for every input sample is done in every layer.According to the self-organization graph calculus model,the information process specifications in every layer and the information transmit specifications between two neighboring layers are given.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第30期28-34,共7页
Computer Engineering and Applications
关键词
多层次自组织
自组图演算
信息粒
容器
复杂认知问题
hierarchical self-organization,self-organization graph calculus,information grains,container,complex recognition problem
