Discussion mechanism based brain storm optimization algorithm

A discussion mechanism based brain storm optimization (DMBSO) algorithm was proposed in order to solve the problem that brain storm optimization (BSO) algorithm is likely to stagnate in the local optima and result in premature convergence. DMBSO used a new mechanism with inter-group discussion and intra-group discussion to replace the process of individual updating in the original BSO algorithm in order to respectively govern the ability of global search and local search. The ability of global search was enhanced at the beginning by linearly decreasing times of inter-group discussion and increasing times of intra-group discussion, while fine search was enhanced in the end to prevent premature convergence. Empirical studies were conducted to evaluate the performances of the DMBSO algorithm for the 10D, 20D, 30D problems of six popular benchmark functions (BFs). Experimental results demonstrate that the DMBSO algorithm can avoid being stagnated in the local optima, more effectively and steadily find the better results than the original BSO algorithm and standard particle swarm optimization (PSO) algorithm, and show stronger robustness with the increasing of BFs' dimension.