Title :
Corrected Subspace Information Criterion for Least Mean Squares Learning
Author_Institution :
Fac. of Math. & Comput. Sci., Huanggang Normal Univ., Huanggang, China
Abstract :
The least mean squares (LMS) algorithm is widely applied in the machine learning community. Corrected subspace information criterion (CSIC) is one of the model selection methods, which is defined on an unbiased estimator of the generalization error-subspace information criterion(SIC). In this paper, we will apply CSIC to select of LMS learning models, it can obtain better results than SIC.
Keywords :
learning (artificial intelligence); least mean squares methods; corrected subspace information criterion; least mean squares learning; machine learning; Computational modeling; Covariance matrix; Kernel; Least squares approximation; Noise; Silicon carbide; Training; corrected subspace information criterion; generalization error; least mean squares algorithm; model selection;
Conference_Titel :
Intelligence Science and Information Engineering (ISIE), 2011 International Conference on
Conference_Location :
Wuhan
Print_ISBN :
978-1-4577-0960-9
Electronic_ISBN :
978-0-7695-4480-9
DOI :
10.1109/ISIE.2011.61
