Title :
Online training algorithm research based on improved weighted LSSVM
Author :
Xing Jianchun ; Wang Ronghao ; Yang Qiliang ; Xiang Zhengrong ; Lu Juliang
Author_Institution :
Eng. Inst. of Eng. Corps, PLA Univ. of Sci. & Technol., Nanjing, China
Abstract :
The traditional weighted LSSVM obtains the weighted value based on predictive error. In this paper, a kind of LSSVM algorithm based on improved weighed function is investigated. The algorithm considers not only the time factor between the training samples and the testing samples but also the similarity factor between them. Furthermore, the algorithm and online idea are combined to model pH neutralization reaction control system. The simulation result indicates that the improvement of the weighed LSSVM is effective. The proposed online training algorithm based on improved weighed LSSVM has the features of good real-time performance and high precision.
Keywords :
chemical engineering computing; chemical reactions; control engineering computing; least squares approximations; pH control; support vector machines; improved weighted LSSVM; online training algorithm; pH neutralization reaction control system; predictive error; similarity factor; time factor; weighted least squares support vector machine; Artificial neural networks; Control systems; Electronic mail; Prediction algorithms; Simulation; Time factors; Training; Online Identification; PH neutralization Reaction; Weighted LSSVM;
Conference_Titel :
Control Conference (CCC), 2010 29th Chinese
Conference_Location :
Beijing
Print_ISBN :
978-1-4244-6263-6
