【摘要】 分析了ＢＰ学习算法中加速因子ε和衰减因子α对算法收敛性和学习精度的影响。在此基础上提出了一种适合于在线控制型神经网络进行训练学习的具有高速收敛性的变因子ＢＰ学习算法，给出了动态调整ε和α的计算公式，通过一个四层前馈感知器型神经网络对一组标准控制序列的学习，证明了该算法的优越性能。更多还原

【Abstract】 The effects of the accelerating factor ε and exponential decay factor α on the convergenceand the learning accuracy of a BP algorithm are discussed.A varying-factor BP learning al-gorithm with fast convergence for the training of on-line controller-type neural network isproposed. The formula for calculating the dynamic adjustment of both factors is given.Through the learning of a set of standard control process by a four-layer feedforward neuralnetwork of the perceptron type,it is shown that the performance of the new algorithm is ex-cellent.更多还原