Title :
Full design matrix designation in orthogonal least squares approximation problems
Author :
Wang, Xunxian ; Brown, David
Author_Institution :
Dept. of Creative Technol., Portsmouth Univ., UK
Abstract :
Based on the forward selection formula, the relationship between the least squares cost function and the correlation between the training data and the regressors is introduced. A rule to design the full design matrix is proposed. A comparison of the experimental data shows that the method is efficient in reducing the complexity of the final approximation model.
Keywords :
correlation methods; least squares approximations; regression analysis; approximation model complexity reduction; forward selection formula; full design matrix designation; kernel regression problem; least squares cost function; orthogonal least squares approximation problems; training data/regressors correlation; Approximation error; Boosting; Cost function; Equations; Intelligent systems; Kernel; Least squares approximation; Least squares methods; Support vector machines; Training data;
Conference_Titel :
Instrumentation and Measurement Technology Conference, 2004. IMTC 04. Proceedings of the 21st IEEE
Print_ISBN :
0-7803-8248-X
DOI :
10.1109/IMTC.2004.1351214
