【摘要】 针对硬件实现RBF神经网络提出了一种亲属优先遗传算法,并用于RBF神经网络对中心值参数c进行优化学习.通过Matlab仿真对非线性函数y=sinc(x)进行逼近,并与BP算法、K-means聚类算法以及标准遗传算法进行比较,实验结果证实了所提出的算法的有效性和适用性,既避免了RBF神经网络的学习陷入局部极小,同时也提高了学习效率,为硬件实现RBF神经网络的片上学习提供了基础.更多还原

【Abstract】 In this paper,a relative priority genetic algorithm is proposed for training the centre parameter c of Radial Basis Function Neural Network. Approximating an nonlinear function y=sinc(x) through Matlab simulation,comparing with BP algorithm,K-means clustering algorithm and the general genetic algorithm,the experimental results show effectiveness and applicability of the proposed algorithm. It not only avoids the learning of neural network away from the local minima,but also improves the learning efficiency.更多还原