We are interested in the solution of Horizontal Linear Complementarity Problems, HLCPs, that is complementarity problems with more variables than equations. Globally metrically regular HLCPs have nonempty solution sets that are stable with respect to “right-hand-side perturbations” of the data, hence are numerically attractive. The main purpose of the paper is to show how the stability or conditioning properties of globally metrically regular HLCPs are preserved by a homotopy framework for solving the HLCP that finds a “stable” direction at each iteration as a local minimizer of a strongly convex quadratic program with linear complementarity constraints, QPCC. Apart from intrinsic interest in numerical solution of HLCPs, this investigation has application in solving horizontal nonlinear complementarity problems and more broadly in the area of mathematical programs with complementarity constraints, MPCCs.
Citation
The Judge Institute of Management, University of Cambridge, Trumpington St, CB2 1AG, UK