A simple pooling model with application in the feed industry and its global solution analysis

Emilie Joannopoulos (1, 2) , François Dubeau (1) , Jean-Pierre Dussault (3) , Mounir Haddou (2) , Candido Pomar (4)

1 UdeS - Département de mathématiques [Sherbrooke]
2 IRMAR - Institut de Recherche Mathématique de Rennes
3 UdeS - Département d'informatique [Sherbrooke]
4 AAFC - Agriculture and Agri-Food

We present a specific pooling model which, while simple, still uses bi-linear objective and constraints. An absolute lower bound is available for this model. We conjecture that any local solution is a global minimizer for this simple instance. We will motivate its usage within a feed industry application. We will then present several experiments to support our conjecture. Since the dimension is in the hundreds, no global solver is able to solve it. Among arguments to support our conjecture is the fact that for smaller instances solvable by global solvers, the solutions computed by local solvers are indeed global.

Mots clés: Global optimization; Convex relaxation; Bilinear models; SDP relaxation; Discretization

Domaines: Optimisation et contrôle [math.OC]

Emilie Joannopoulos, François Dubeau, Jean-Pierre Dussault, Mounir Haddou, Candido Pomar. A simple pooling model with application in the feed industry and its global solution analysis. 2017. ⟨hal-01470800⟩