论文部分内容阅读
为了提高模糊神经网络(FNN)的收敛速度和泛化能力,提出一种基于混合梯度下降算法(HG)的模糊神经网络(HG-FNN).HG-FNN通过设计FNN参数调整过程的自适应学习率,利用链式法则获取FNN参数学习过程的梯度,在实现FNN参数自校正的同时,给出HG-FNN的收敛性证明,保证HG-FNN的收敛速度和泛化能力.最后,将所设计的HG-FNN应用于非线性系统建模与污水处理过程关键水质参数预测,实验比较结果显示,HG-FNN不仅具有较快的收敛速度,而且具有较好的泛化能力.
In order to improve the convergence speed and generalization ability of FNN, a fuzzy neural network (HG-FNN) based on Hybrid Gradient Descent Algorithm (HG) is proposed.HG-FNN is designed by designing adaptive learning of FNN parameter adjustment process Rate, the chain rule is used to obtain the gradient of FNN parameter learning process. At the same time, the convergence of HG-FNN is proved and the convergence rate and generalization ability of HG-FNN are given.Finally, HG-FNN is applied in the prediction of key water quality parameters in nonlinear system modeling and sewage treatment. The experimental results show that HG-FNN not only has faster convergence rate but also has better generalization ability.