一种RBF神经网络的自适应学习算法

摘要

A new self-adaptive learning model of RBF neural network was established successfully.The model does not beed to confirm the center position and the quantity of the hidden layer nobes beforehand, but adds or removes the quantity of the nobes according to corresponding additive strategy and removing strategy in the process of recognition.The finally formed network has a simple structure, high accuracy and better adaptive ability.%建立了一种RBF神经网络的自适应学习模型.该模型事先不需要确定隐层节点的中心位置和数量,而是在学习过程中,根据相应的添加策略和删除策略,自适应地增加或减少隐层节点的数量.最终形成的网络不仅结构简单,精度高,而且具有较好的泛化能力.