论文部分内容阅读
标准的或许多改进的BP算法大都采用均方误差估计器,致使学习易于陷入局部极小、收敛速度慢和对初始权敏感等。本文针对这些缺点提出了较好的算法,仿真结果证实了核算法的正确性。
Most of the standard or many improved BP algorithms use mean square error estimators, which make learning easy to fall into local minima, slow convergence and sensitivity to initial rights. In this paper, we propose a good algorithm for these shortcomings. The simulation results confirm the correctness of the algorithm.