论文部分内容阅读
生成模糊系统传统方法的工作量往往随输入变量数的增长而爆炸性地增加，用于抽取模糊规则的神经网络的规模迅速地增加且能量的极小值点也迅速地增多。针对这一问题，本文发展了一种新的模糊系统生成方法：将复杂系统的模糊输入、输出关系分解成简单的模糊输入、输出关系叠加，采用了一种新的网络优化的方法──基于浮点编码的遗传算法（Floatcodingbasedgeneticalgorithm，FGA）来生成该系统．FGA克服了BP算法的网络麻痹、收敛于能量局域最小点、阈值函数连续可导性限制等缺点；同时也解决了一般遗传算法不符合于生物演化的规律，限定搜索范围等问题。FGA实现了在实数空间中不同尺度上的同时公平的搜索。由于化繁为简的策略以及FGA的采用，生成系统的工作量并不随系统的输入变量数的增长而剧增，也不会出现网络收敛于能量局域最小点的情况。因此本方法可以用以产生实用的复杂多输入模糊系统。
The workload of traditional methods for generating fuzzy systems often increases explosively with the number of input variables. The scale of neural networks for extracting fuzzy rules increases rapidly and the minimum value of energy increases rapidly as well. In order to solve this problem, a new method for generating fuzzy systems is developed. The fuzzy input and output relations of complex systems are decomposed into simple fuzzy inputs and the output relations are superposed. A new method of network optimization Float coding genetic algorithm (Floatcodingbasedgeneticalgorithm, FGA) to generate the system. FGA overcomes the network paralysis of BP algorithm and converges to the shortest point in energy domain and the continuous conductivity limit of threshold function. FGA also solves the problems that general genetic algorithm does not accord with the law of biological evolution and limits the scope of search. The FGA achieves simultaneous and fair search on different scales in real space. Due to the simplistic strategy and the adoption of FGA, the workload of the generated system does not increase sharply with the increase of the number of input variables of the system, nor does the situation that the network converges to the minimum point in the energy domain. Therefore, this method can be used to generate a practical complex multi-input fuzzy system.