Cooperated Bayesian algorithm for distributed scheduling problem

Abstract

This paper presents a new distributed Bayesian optimization algorithm (BOA) to overcome the efficiency problem when solving NP scheduling problems. The proposed approach integrates BOA into the co-evolutionary schema, which builds up a concurrent computing environment. A new search strategy is also introduced for local optimization process. It integrates the reinforcement learning (RL) mechanism into the BOA search processes, and then uses the mixed probability information from BOA (post-probability) and RL (pre-probability) to enhance the cooperation between different local controllers, which improves the optimization ability of the algorithm. The experiment shows that the new algorithm does better in both optimization (2.2 %) and convergence (11.7 %), compared with classic BOA.

Keywords

statistic optimization, distributed scheduling, Bayesian networks, data mining