用综合法优化前向神经网络结构

在神经网络研究中,如何确定神经网的结构是一个重要的研究方向.提出了一种通用的确定前向神经网络结构的自适应方法,即先用动态增长法快速训练网络拓扑结构及权值至满足给定的误差为止,然后用遗传算法(GA)对训练好的网络剪枝.实验表明,算法具有较好的通用性和可扩展性,收敛速度较快,对进一步的数据挖掘具有重要的意义.

Optimal feed-forward neural network architecture using synthesis method

On neural network researching,how to determining the neural network architecture is an important researching branch.An adaptive approach of determining topology structure of feed-forward neural network is presented in this paper.At first,the network is trained by a dynamic increasing method until the error is satisfied.And then the trained network is pruned by Genetic Algorithm(GA).It is shown by experiments that the approach this paper given is general and expending,and its convergence speed is fast.Moreover,it is of momentous significance for further researching on data mining.