Regularized multi-view least squares twin support vector machines

Abstract

Regularized least squares twin support vector machines are a new nonparallel hyperplane classifier, which can lead to simple and fast algorithms for generating binary classifiers by replacing inequality constraints with equality constraints and implementing the structural risk minimization principle in twin support vector machines. Multi-view learning is an emerging direction in machine learning which aims to exploit distinct views to improve generalization performance from multiple distinct feature sets. Experimental results demonstrate that our proposed methods are effective.