Quadratic and convex minimax classification problems. (English) Zbl 1279.90125
Summary: When there are two classes whose mean vectors and covariance matrices are known, G. R. G. Lanckriet, L. El Ghaoui, C. Bhattacharyya and {M. Jordan} [J. Mach. Learn. Res. 3, No. 3, 555-582 (2003; Zbl 1084.68657)] consider the Linear Minimax Classification (LMC) problem and they propose a method for solving it. In this paper we first discuss the Quadratic Minimax Classification (QMC) problem, which is a generalization of LMC. We show that QMC is transformed to a parametric Semidefinite Programming (SDP) problem. We further define the Convex Minimax Classification (CMC) problem. Though the two problems are generalizations of LMC, we prove that solutions of these problems can be obtained by solving LMC.
MSC:
90C20 | Quadratic programming |
90C25 | Convex programming |
90C47 | Minimax problems in mathematical programming |