Section: New Results
DAMS: Distributed Adaptive Metaheuristic Selection
Participants : B. Derbel, S. Verel.
In this work, we design a new Distributed Adaptive Metaheuristic Selection (DAMS) scheme. DAMS is dedicated to adaptive optimization in distributed environments. Given a set of metaheuristics, the goal of DAMS is to coordinate their local execution on distributed nodes in order to optimize the global performance of the distributed system. DAMS is based on three-layer architecture allowing nodes to decide distributively what local information to communicate, and what metaheuristic to apply while the optimization process is in progress. Within this context, we specialize DAMS by describing a particular instantiation called Select Best and Mutate (SBM). Its is a simple, yet efficient, adaptive distributed algorithm using an exploitation component allowing nodes to select the metaheuristic with the best locally observed performance, and an exploration component allowing nodes to detect the metaheuristic with the actual best performance. SBM features are analyzed from both a parallel and an adaptive point of view, and its efficiency is demonstrated through experimentations and comparisons with other adaptive strategies (sequential and distributed).