Section: New Results

Search and modelling

Participants : Eric Monfroy, Thierry Petit.

• In the context of autonomous search, [21] deals with the problem of automatically tuning a search strategy (i.e., variable value selection). For this purpose it uses so called choice functions which provide an evaluation of a strategy in term of a set of indicators. [36] and [31] go one step further by providing tuning and adaptation facilities at the level of the different components of a constraint solver.

• Using the MiniZinc modeling language, [32] shows how to model and solve the portfolio selection problem with constraint programming. Since more than ten year constraints for which the set of solutions can be matched to the language accepted by an automaton were introduced in many solvers (e.g., Choco, Gecode, SICStus). [40] describes an interface for describing such constraints in a more convenient way.

• Many discrete optimization problems have constraints on the objective function. Being able to represent such constraints is fundamental to deal with many real world industrial problems. In this work, we go one step further in the concept of topologically concentrate high values in a sequence of cost variables. We refine the work we previously published in CP2012 thanks to three generalizations of the focus constraint. We experiment successfully the technique in scheduling, round-robin and musical benchmarks. This work has been published at IJCAI 2013 [37] .