论文部分内容阅读
提出了新的遗传算法优化设计前向神经网络的结构和权重矢量。这种新方法的创新在于:二值码串和实值码串的混合编码方法即保留了传统遗传算法的优点,又具有遗传编程和遗传策略的优点;结合遗传算子和SolisandWets算法生成后代的方法丰富了遗传搜索空间的多样性,加快了遗传算法的收敛速度;对混合编码码串的动态参数编码方法提高了优化精度。
Proposed a new genetic algorithm to optimize the design of forward neural network structure and weight vector. The innovation of this new method lies in that the hybrid encoding method of binary code string and real value code string retains the advantages of the traditional genetic algorithm and has the advantages of genetic programming and genetic strategy; combines the genetic operator and the SolisandWets algorithm to generate the offspring The method enriches the diversity of the genetic search space and accelerates the convergence rate of the genetic algorithm. It improves the precision of the dynamic parameter coding method of hybrid code string.