一种实用的克服局部极小的BP算法研究

A USEFUL BP ALGORITHM TO OVERCOME LOCAL MINIMA

  • 摘要: BP算法由于其神经元输出函数为Sigmoid函数,因此是一个非线性优化问题,不可避免地会出现局部极小.本文提供了一种改进的学习算法,提出判断局部最小的规则,能后引入冲量函数,使BP网络能够通过判断输出节点的输出误差来修改学习率,使误差函数在其梯度方向上出现大的跳跃,从而跳出局部极小.

     

    Abstract: BP algorlthm is actually a nonlinear optimal problem and produces inevitably Iocal minima.The paper proposes the rules to judge local minima and introduces momentum function in BP algorithm. Thealgorithm may iump local minima,Simulation result is given to show the effectiveness and the usage of the algorithm.

     

/

返回文章
返回