HIPCUT - A hybrid interior-point cutting-plane method for integer

Programming

Alexander Engau

To solve the continuous relaxations of discrete optimization problems,

we integrate a cutting-plane scheme into the general framework of an

infeasible interior-point algorithm. Instead of adding cuts only after

finding approximate solutions to successive relaxations, separating

inequalities are also added and removed already during the solution

process based on feasibility indicators at intermediate iterates.

Preliminary computational results are reported for instances of the

max-cut problem.