INFEASIBLE FULL NEWTON-STEP INTERIOR-POINT METHOD FOR LINEAR COMPLEMENTARITY PROBLEMS

Abstract

In this paper we consider an Infeasible Full Newton-step Interior-Point Method (IFNS-IPM) for monotone Linear Complementarity Problems (LCP). The method does not require a strictly feasible starting point. In addition, the method avoids calculation of the step size and instead takes full Newton-steps at each iteration. Iterates are kept close to the central path by suitable choice of parameters. The algorithm is globally convergent and the iteration bound matches the best known iteration bound for these types of methods.

Keywords

• infeasible full Newton-step interior-point method
• linear complementarity problem
• iteration bound