Abstract: Mixed-integer formulation of the problem of minimization of empirical risk is considered. Some possibilities of decision of the continuous relaxation of this problem are analyzed. Comparison of the proposed continuous relaxation with a similar SVM problem is performed too.

Keywords: cluster, decision rule, discriminant function, linear and non-linear programming, non-smooth optimization

ACM Classification Keywords: G.1.6 Optimization - Gradient methods, I.5 Pattern Recognition; I.5.2 Design Methodology - Classifier design and evaluation

Link:

MINIMIZATION OF EMPIRICAL RISK IN LINEAR CLASSIFIER PROBLEM

Yurii I. Zhuravlev, Yury Laptin, Alexander Vinogradov

http://foibg.com/ibs_isc/ibs-16/ibs-16-p01.pdf