摘要
Slow convergence of back-propagation (BP) algorithm is a limiting factor in its practical applications. A new learning algorithm which can adaptively adjust its learning rate on the basis of gradient information of the error function is put forward. its convergence performance is also tested by the XOR problem compared with the standard BP algorithm.
Slow convergence of back-propagation (BP) algorithm is a limiting factor in its practical applications. A new learning algorithm which can adaptively adjust its learning rate on the basis of gradient information of the error function is put forward. its convergence performance is also tested by the XOR problem compared with the standard BP algorithm.
