模糊进化神经网络理论与技术框架

Framework of Fuzzy Evolutionary Neural Network Theory and Technology

分析目前进化算法、神经网络和模糊逻辑等理论技术及应用研究存在的问题 ,以及目前这些智能理论融合技术和神经网络自动设计研究现状 ,研究了现有智能理论技术相互之间包含的智能行为及属性 ,提出了模糊进化神经网络理论。其中 ,给出基本概念、研究目的、内容和框架体系 ,详细描述了模糊进化计算、模糊进化神经网络和进化模糊系统从模型初始化、参数自适应到网络自动设计、模型评价等核心技术及其解决思路和实现算法。对于模糊进化计算 ,提出了其控制参数模糊初选 ,模糊自适应进化 ,优化算法及其结果的模糊评价等方法 ;对于模糊进化神经网络 ,提出了其全自动设计 ,控制参数模糊初选 ,模糊自适应进化以及网络模型的模糊评价与选择等方法 ;对于进化模糊系统 。

The paper analyzes the existing defects of the theory and application of evolutionary algorithm, neural network and fuzzy logic, the present fusing technologies of intelligent theory and automatic design method of neural network, researches the mutually contained intelligent action and properties of these intelligent theories. Then it proposes the fuzzy evolutionary neural network theory. In the theory, its definition, goal, content and framework system are given, the key serial technologies and their basic ideas and algorithm for fuzzy evolutionary algorithm (FEA),fuzzy evolutionary neural network (FENN) and evolutionary fuzzy system (EFS) from model initialization, self adaptation to automatic design and model evaluation are described. Firstly the approaches of initial selection and self adaptation of control parameters and evaluation means for FEA are put forward. Then the approaches of automatic design, fuzzy initialization and self adaptive evolution of FENN parameters and evaluation of network model are researched for FENN. Finally the parameters optimization approaches of the fuzzy system and fuzzy neural network based on FEA are researched in EFS.