Improvements on ν-Twin Support Vector Machine

Khemchandani, Reshma

doi:10.1016/j.neunet.2016.03.011

Improvements on ν-Twin Support Vector Machine

2016-07-01

31

PRE

AI

K

Khemchandani, Reshma *

P

Pooja Saigal

DOI：10.1016/j.neunet.2016.03.011

原文链接

原文求助

摘要

En 中文

In this paper, we propose two novel binary classifiers termed as Improvements on nu-Twin Support Vector Machine: I nu-TWSVM and I nu-TWSVM (Fast)'' that are motivated by nu-Twin Support Vector Machine (nu-TWSVM). Similar to nu-TWSVM, I nu-TWSVM determines two nonparallel hyperplanes such that they are closer to their respective classes and are at least rho distance away from the other class. The significant advantage of I nu-TWSVM over nu-TWSVM is that I nu-TWSVM solves one smaller-sized Quadratic Programming Problem (QPP) and one Unconstrained Minimization Problem (UMP); as compared to solving two related QPPs in nu-TWSVM. Further, I nu-TWSVM (Fast) avoids solving a smaller sized QPP and transforms it as a unimodal function, which can be solved using line search methods and similar to I nu-TWSVM, the other problem is solved as a UMP. Due to their novel formulation, the proposed classifiers are faster than nu-TWSVM and have comparable generalization ability. I nu-TWSVM also implements structural risk minimization (SRM) principle by introducing a regularization term, along with minimizing the empirical risk. The other properties of I nu-TWSVM, related to support vectors (SVs), are similar to that of nu-TWSVM. To test the efficacy of the proposed method, experiments have been conducted on a wide range of UCI and a skewed variation of NDC datasets. We have also given the application of I nu-TWSVM as a binary classifier for pixel classification of color images. (C) 2016 Elsevier Ltd. All rights reserved.

Keyword:

Twin support vector machines

Machine learning

Unconstrained optimization

Image pixel classification

Improvements on ν-Twin Support Vector Machine

摘要

AI总结

期刊

机构