TY - JOUR
T1 - Approximate model selection for large scale LSSVM
AU - Ding, Lizhong
AU - Liao, Shizhong
PY - 2011
Y1 - 2011
N2 - Model selection is critical to least squares support vector machine (LSSVM). A major problem of existing model selection approaches of LSSVM is that the inverse of the kernel matrix need to be calculated with O(n3) complexity for each iteration, where n is the number of training examples. It is prohibitive for the large scale application. In this paper, we propose an approximate approach to model selection of LSSVM. We use multilevel circulant matrices to approximate the kernel matrix so that the fast Fourier transform (FFT) can be applied to reduce the computational cost of matrix inverse. With such approximation, we first design an efficient LSSVM algorithm with O(n log(n)) complexity and theoretically analyze the effect of kernel matrix approximation on the decision function of LSSVM. We further show that the approximate optimal model produced with the multilevel circulant matrix is consistent with the accurate one produced with the original kernel matrix. Under the guarantee of consistency, we present an approximate model selection scheme, whose complexity is significantly lower than the previous approaches. Experimental results on benchmark datasets demonstrate the effectiveness of approximate model selection.
AB - Model selection is critical to least squares support vector machine (LSSVM). A major problem of existing model selection approaches of LSSVM is that the inverse of the kernel matrix need to be calculated with O(n3) complexity for each iteration, where n is the number of training examples. It is prohibitive for the large scale application. In this paper, we propose an approximate approach to model selection of LSSVM. We use multilevel circulant matrices to approximate the kernel matrix so that the fast Fourier transform (FFT) can be applied to reduce the computational cost of matrix inverse. With such approximation, we first design an efficient LSSVM algorithm with O(n log(n)) complexity and theoretically analyze the effect of kernel matrix approximation on the decision function of LSSVM. We further show that the approximate optimal model produced with the multilevel circulant matrix is consistent with the accurate one produced with the original kernel matrix. Under the guarantee of consistency, we present an approximate model selection scheme, whose complexity is significantly lower than the previous approaches. Experimental results on benchmark datasets demonstrate the effectiveness of approximate model selection.
KW - Least squares support vector machine
KW - Matrix approximation
KW - Model selection
KW - Multilevel circulant matrices
UR - http://www.scopus.com/inward/record.url?scp=84876851009&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:84876851009
SN - 1532-4435
VL - 20
SP - 165
EP - 180
JO - Journal of Machine Learning Research
JF - Journal of Machine Learning Research
T2 - 3rd Asian Conference on Machine Learning, ACML 2011
Y2 - 13 November 2011 through 15 November 2011
ER -