【作者】 姜燕； 何青； 童调生；

【Author】 Yan Jiang Qing He Tiaosheng Tong (College of Electrical & Information Engineering, Hunan University Changsha 410082)

【机构】 湖南大学电气与信息工程学院；

【摘要】 本文研究了未知但有界误差下神经网络的学习算法问题,在分析传统学习算法的基础上,提出了一种最优定界椭球学习方法,该方法具有收敛速度快,精度高等优点。通过对NARMA模型参数估计的仿真结果表明此算法的有效性。更多还原

【Abstract】 This paper presents the problem of the learning algorithm in neural networks under unknown but bounded (UBB) error. Based on the analysis of the traditional methods, a new optimal bounding ellipsoid learning algorithm is proposed. This algorithm has some better properties, such as better convergence speed and accuracy. The validity of the algorithm is proved through the simulation results of the parameter estimation of NARMA model.更多还原