Authors: Juan Lin; Yiwen Zhong
Addresses: College of Computer and Information Science, Fujian Agriculture and Forestry University, Fuzhou 350002, China ' College of Computer and Information Science, Fujian Agriculture and Forestry University, Fuzhou 350002, China
Abstract: Traditional list-based threshold-accepting (LBTA) algorithm is similar with simulated annealing (SA) algorithm, depends on an intense local search method, and utilises a list filling procedure with threshold values to search the solution space effectively. Inspired by the learning ability of particle swarm optimisation (PSO), multi-agent LBTA (MLBTA) involves the learning knowledge to guide its sampling, explores the solution space in a co-evolution mode. Compare with multi-agent SA (MSA) algorithm adapting the same local search version, MLBTA incorporates a dynamic list of threshold values which is adapted according to the topology of the solution space and tunes only one parameter. Dispense with sophisticated parameters as MSA, MLBTA balances the intensification and diversification iteratively. Computational results on functions optimisation and protein structure prediction (PSP) problems show that MLBTA algorithm achieves better or comparable performances with MSA.
Keywords: list-based algorithms; threshold-accepting algorithms; multi-agent systems; MAS; agent-based systems; particle swarm optimisation; PSO; numerical optimisation; local search; simulated annealing; intensification; diversification; functions optimisation; protein structure prediction; PSP.
International Journal of Computing Science and Mathematics, 2015 Vol.6 No.5, pp.501 - 509
Received: 24 Jun 2015
Accepted: 11 Jul 2015
Published online: 10 Nov 2015 *