Unbounded convex sets for non-convex mixed-integer quadratic programming (Q2436643)

This article considers the non-convex mixed-integer quadratic programming problem which is converted into a problem with a linear objective function but non-linear constraints, with the view of deriving several classes of facet-defining inequalities. The article begins with a brief introduction to quadratic programming and the necessary background theorems and notation. This is followed by the definition of the convex sets and their properties, such as dimension, complexity and symmetries. Classes of new valid inequalities are then described and proven, firstly for the easier case where all variables are continuous, secondly for the case of all-integer variables and finally the general mixed-integer case. Several properties of the convex sets associated with unconstrained, non-convex, mixed-integer quadratic problems are outlined and proven in this very interesting paper, giving rise to new cutting planes which can be used to enhance algorithms for the exact solution of quadratic programs.