DocumentCode :
2232854
Title :
Kernel matrix approximation for learning the kernel hyperparameters
Author :
Fauvel, Mathieu
Author_Institution :
INRA, Univ. of Toulouse, Toulouse, France
fYear :
2012
fDate :
22-27 July 2012
Firstpage :
5418
Lastpage :
5421
Abstract :
The selection of kernel hyperparameters is addressed in this article. The proposed method is based on the approximation of an empirical ideal kernel matrix using three measures of similarity between matrices. The conventional kernel alignment, the Frobenius distance and the correlation between matrices are investigated. A quadratic gradient optimization is proposed to find the set of hyperparameters maximizing the similarity between the empirical ideal kernel matrix and the sample kernel matrix. The Gaussian kernel is used in this article for numerical experiments. Classification of several real data sets are performed. In terms of classification accuracies and processing time, results show that the proposed approach is effective for tuning the kernel hyperparameters.
Keywords :
Gaussian processes; matrix algebra; optimisation; pattern classification; Frobenius distance; Gaussian kernel; empirical ideal kernel matrix approximation; hyperparameter maximization; kernel alignment; kernel hyperparameter learning; kernel hyperparameter selection; matrix similarity; quadratic gradient optimization; Accuracy; Approximation methods; Educational institutions; Kernel; Optimization; Remote sensing; Support vector machines;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Geoscience and Remote Sensing Symposium (IGARSS), 2012 IEEE International
Conference_Location :
Munich
ISSN :
2153-6996
Print_ISBN :
978-1-4673-1160-1
Electronic_ISBN :
2153-6996
Type :
conf
DOI :
10.1109/IGARSS.2012.6352381
Filename :
6352381
Link To Document :
بازگشت