Optimality-based domain reduction for inequality-constrained NLP and MINLP problems (Q2182853)

In this paper, first-order optimality conditions for nonlinear programming (NLP) and mixed-integer nonlinear programming (MINLP) problems are developed. Based on these conditions, the authors present three separate optimality-based reduction algorithms for unconstrained, one-constraint, and multi-constraint problems that may be purely continuous or involve integer variables. By dropping redundant constraints at each node of the branch-and-bound tree, the reduction algorithms are flexibly applied to equality-constrained problems. An efficient implementation of the presented algorithms in the branch-and-bound solver BARON is given. The computations illustrate that these algorithms often reduce the solution time and total number of nodes for continuous problems, but they are less effective for mixed-integer programs.