Abstract:

In this seminar we consider convex Mixed Integer Nonlinear Programming (MINLP), i.e., mixed integer optimization problems with convex objective function and constraints. In the first part of talk we introduce a Quadratic Outer Approximation framework for convex MINLP and we discuss several strategies to determine convex quadratic underestimations of convex functions. Then, we present a QP/NLP-based Branch-and-Bound scheme obtained by generalizing the LP/NLP-based Branch-and-Bound introduced by Quesada and Grossmann in their 1992 seminal paper. Preliminary promising computational results will be presented and discussed.