论文部分内容阅读
针对普通ＢＰ算法学习周期长的不足，根据网络误差表面形状，文章在学习算法中引入模糊理论，动态地在线调整网络的学习步长η及动量因子α。仿真结果表明，模糊理论的引入改善了学习算法的性能。
Aiming at the problem of long learning cycle of ordinary BP algorithm, according to the shape of network error surface, the article introduces fuzzy theory into learning algorithm to dynamically adjust network learning step η and momentum factor α dynamically. Simulation results show that the introduction of fuzzy theory improves the performance of learning algorithm.