A new algorithm is proposed to solve the 'flat spot' problem in back-propagation neural networks by magnifying the gradient function. Simulation results show that, in terms of the convergence rate and the perc...
详细信息
A new algorithm is proposed to solve the 'flat spot' problem in back-propagation neural networks by magnifying the gradient function. Simulation results show that, in terms of the convergence rate and the percentage of global convergence, the new algorithm consistently outperforms other traditional methods.
暂无评论