Section: New Results

Meta Heuristics

Participants : Alejandro Reyes Amaro, Eric Monfroy, Florian Richoux, Charlotte Truchet.

• The aim is to develop and implement new algorithmical methods for constraint problems on massively parallel machines. We also conduct more theoretical studies about the parallelization of constraint problems. This year, we proposed a fairly sharp model to predict parallel speed-ups one can expect while parallelizing by a multi-walk parallel scheme any Las Vegas algorithm by just studying the distribution of sequential run-times [41] . This model shows a divergence of only 20% when predicting speed-ups over 256 cores, on very different benchmarks.

• To evaluate the scalability and parallelization of local search algorithms for SAT, [23] presents a statistical method based on the analysis of the runtime behavior of its sequential version.

• [26] and [26] deals with the use of metaheuristics for solving the resource constrained scheduling problem and the set covering problems.