【作者】 李玉惠； 李勃； 张天贵；

【Author】 Li Yuhui Li Bo Automation Dpartment,Kunming University of Science and Technology,Kunming,650093 Zhang Tiangui Yunnan Chemical Plant of Natural Gas,Suifu,657800

【机构】 昆明理工大学自动化系； 云南天然气化工厂；

【摘要】 本文针对神经元网络定限值控制存在训练时间长的不足,首次提出一种反偏差方法,即为加快系统初始响应,又不致使超调过大,在输出神经元阈值中引入反偏差函数,在不改变网络学习算法的情况下改进网络的学习状态,缩短网络训练时间,提高网络的收敛速度,该方法具有普遍实用性,为神经网络控制和应用提供了一种新思路．更多还原

【Abstract】 This paper, being directed against the drawback that the training time for the Constant-limiuing control of the neural networks is longer, suggests a way of the inverse deviation That is to accelerate initial response without making modulation larger. Being added inverse deviation function in output ncurals threshold without changing the condition of learning algorithem of the neural networks, the learning condition of the neural networks can be improved,its training time can be reduced and its convergence speed can be raised The way can be wideiy used. It gives an new idea in the use of the neural networks control.更多还原