WSEAS Transactions on Mathematics
Print ISSN: 1109-2769, E-ISSN: 2224-2880
Volume 15, 2016
The Influences of Smooth Approximation Functions for SPTSVM
Authors: ,
Abstract: The recently proposed smooth projection twin support vector machine(SPTSVM) gains a good generalization ability and is suitable for many binary classification problems. But we know that different smooth approximation functions may bring different classification accuracies. In order to study the influence of smooth approximation functions for SPTSVM, in this paper, we first overview eight known smooth approximation functions and describe their differentiability and error ranges by five lemmas and one theorem. Then, we perform a series of comparative experiments on classification accuracy and running time by using SPTSVM with Newton-Armijo method on 10 UCI datasets and 6 NDC datasets. From experiment results, we can get a choice order of the eight approximation functions in generally.
Search Articles
Pages: 241-250
WSEAS Transactions on Mathematics, ISSN / E-ISSN: 1109-2769 / 2224-2880, Volume 15, 2016, Art. #23