本文選題：模糊神經(jīng)網(wǎng)絡(luò) + 混合梯度��； 參考：《控制與決策》2017年09期

【摘要】：為了提高模糊神經(jīng)網(wǎng)絡(luò)(FNN)的收斂速度和泛化能力,提出一種基于混合梯度下降算法(HG)的模糊神經(jīng)網(wǎng)絡(luò)(HG-FNN).HG-FNN通過設(shè)計FNN參數(shù)調(diào)整過程的自適應(yīng)學(xué)習(xí)率,利用鏈?zhǔn)椒▌t獲取FNN參數(shù)學(xué)習(xí)過程的梯度,在實現(xiàn)FNN參數(shù)自校正的同時,給出HG-FNN的收斂性證明,保證HG-FNN的收斂速度和泛化能力.最后,將所設(shè)計的HG-FNN應(yīng)用于非線性系統(tǒng)建模與污水處理過程關(guān)鍵水質(zhì)參數(shù)預(yù)測,實驗比較結(jié)果顯示,HG-FNN不僅具有較快的收斂速度,而且具有較好的泛化能力.
[Abstract]:In order to improve the convergence speed and generalization ability of fuzzy neural networks (FNNs), a hybrid gradient descent algorithm (HG-based) based on fuzzy neural networks (FNNNs) is proposed. The adaptive learning rate of the FNN parameter adjustment process is designed. The gradient of FNN parameter learning process is obtained by using chain rule. The self-tuning of FNN parameters is realized, and the convergence proof of HG-FNN is given to ensure the convergence speed and generalization ability of HG-FNN. Finally, the designed HG-FNN is applied to the modeling of nonlinear systems and the prediction of key water quality parameters in the process of sewage treatment. The experimental results show that HG-FNN not only has a faster convergence rate, but also has a better generalization ability.
【作者單位】： 北京工業(yè)大學(xué)電子信息與控制工程學(xué)院;北京工業(yè)大學(xué)計算智能與智能系統(tǒng)北京市重點實驗室;
【基金】：國家自然科學(xué)基金項目(61533002,61622301) 北京市自然科學(xué)基金項目(4172005) 科技部水專項(2017ZX07104)
【分類號】：TP183

【相似文獻】

相關(guān)期刊論文 前10條

1 邢進生,安凱,萬百五;模糊神經(jīng)網(wǎng)絡(luò)的記憶[J];西安交通大學(xué)學(xué)報;2001年02期

2 王旭e，

本文編號：1791945