Application of Lagrangian Twin Support Vector Machines for Classification

In this paper a new iterative approach is proposed for solving the Lagrangian formulation of twin support vector machine classifiers. The main advantage of our method is that rather than solving a quadratic programming problem as in the case of the standard support vector machine the inverse of a matrix of size equals to the number of input examples needs to be determined at the very beginning of the algorithm. The convergence of the algorithm is stated. Experiments have been performed on a number of interesting datasets. The predicted results are in good agreement with the observed values clearly demonstrates the applicability of the proposed method.