论文部分内容阅读
本文介绍了把线性不可分问题分解为一系列线性可分子问题、对线性不可分问题进行求解的网络分解重组算法．还证明了该算法的收敛性．实例研究表明：该算法不仅可以得到神经网络的隐层空间目标和隐层单元数，而且提高了对线性不可分问题的求解速度，因此是一个非常有效的神经网络训练算法．
This paper introduces a network decomposition and reorganization algorithm that decomposes the linear inseparable problem into a series of linear and separable linear problems and solves the linear inseparable problem. The convergence of the algorithm is also proved. The case study shows that this algorithm can not only obtain the hidden layer space target and hidden layer unit number of neural network, but also improve the speed of solving the linear inseparable problem. So it is a very effective neural network training algorithm.