Mean field variational framework for integer optimization

Abstract

A principled method to obtain approximate solutions of general constrained integer optimization problems is introduced. The approach is based on the calculation of a mean field probability distribution for the decision variables which is consistent with the objective function and the constraints. The original discrete task is in this way transformed into a continuous variational problem. In the context of particular problem classes at small and medium sizes, the mean field results are comparable to those of standard specialized methods, while at large sized instances is capable to find feasible solutions in computation times for which standard approaches can't find any valuable result. The mean field variational framework remains valid for widely diverse problem structures so it represents a promising paradigm for large dimensional nonlinear combinatorial optimization tasks.