| 一种基于混合梯度下降算法的模糊神经网络设计及应用 |
| |
| 引用本文： | 韩红桂,林征来,乔俊飞.一种基于混合梯度下降算法的模糊神经网络设计及应用[J].控制与决策,2017,32(9):1635-1641. |
| |
| 作者姓名： | 韩红桂 林征来 乔俊飞 |
| |
| 作者单位： | 1. 北京工业大学电子信息与控制工程学院,北京100124;2. 北京工业大学计算智能与智能系统北京市重点实验室,北京100124,1. 北京工业大学电子信息与控制工程学院,北京100124;2. 北京工业大学计算智能与智能系统北京市重点实验室,北京100124,1. 北京工业大学电子信息与控制工程学院,北京100124;2. 北京工业大学计算智能与智能系统北京市重点实验室,北京100124 |
| |
| 基金项目： | 国家自然科学基金项目(61533002,61622301)；北京市自然科学基金项目(4172005)；科技部水专项(2017ZX07104). |
| |
| 摘 要： | 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.
|
| 关 键 词： | 模糊神经网络 混合梯度 自适应学习率 非线性系统建模 |
Design and application of hybrid gradient descent-based fuzzy neural network |
| |
| HAN Hong-gui,LIN Zheng-lai and QIAO Jun-fei.Design and application of hybrid gradient descent-based fuzzy neural network[J].Control and Decision,2017,32(9):1635-1641. |
| |
| Authors: | HAN Hong-gui LIN Zheng-lai and QIAO Jun-fei |
| |
| Affiliation: | 1. College of Electronic Information & Control Engineering,Beijing University of Technology,Beijing 100124,China;2. Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing University of Technology,Beijing 100124,China,1. College of Electronic Information & Control Engineering,Beijing University of Technology,Beijing 100124,China;2. Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing University of Technology,Beijing 100124,China and 1. College of Electronic Information & Control Engineering,Beijing University of Technology,Beijing 100124,China;2. Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing University of Technology,Beijing 100124,China |
| |
| Abstract: | To improve the convergence speed and generalization ability of the fuzzy neural network(FNN), a fuzzy neural network, based on the hybrid gradient(HG) descent algorithm, is proposed in this paper. This HG-FNN can obtain the adaptive learning rate of the parameter adjustment process. Then, the chain rule is used to calculate the gradient descent of the learning process to adjust the parameters of FNN. Meanwhile, the convergence proof of HG-FNN is given in details to ensure the convergence speed and the precision of FNN. Finally, the proposed HG-FNN is used to model the nonlinear systems and predict the effluent qualities of wastewater treatment process. The results show that the proposed HG-FNN owns faster convergence speed, as well as with suitable generalization ability than other FNNs. |
| |
| Keywords: | |
|
| 点击此处可从《控制与决策》浏览原始摘要信息 |
|
点击此处可从《控制与决策》下载全文 |
|
