论文部分内容阅读
本文采用一类正交多项式集合作为神经元的激励函数,构成一个正交多项式基神经网络。网络的拓扑结构和相应的正交多项式基在学习训练的过程中确定,网络的权值经最小二乘算法得到,避免了局部极值问题。仿真结果表明,本文提出的方法是可行和有效的。
In this paper, we use a set of orthogonal polynomial sets as the excitation function of neurons to form an orthogonal polynomial basis neural network. The topological structure of the network and the corresponding orthogonal polynomial basis are determined during the learning and training. The weight of the network is obtained by the least-squares algorithm, which avoids the problem of local extremum. Simulation results show that the proposed method is feasible and effective.