Dual approaches for a specific class of integer nonlinear programming problems

In this work, we propose a strategy for computing valid lower bounds for a specific class of integer nonlinear programming problems, that includes integer quadratic programming problems. This strategy is used within a branch-and-bound scheme. Experimental results for randomly generated instances show that, in the quadratic case, the devised branch-and-bound method compares favorably to the MIQP solver of CPLEX 12.6 when the number of constraints is small.