| Abstract: | Firefly algorithm(FA) is a recently-proposed swarm intelligence technique. It has shown good performance in solving various optimization problems. According to the standard firefly algorithm and most of its variants, a firefly migrates to every other brighter firefly in each iteration. However, this method leads to defects of oscillations of positions, which hampers the convergence to the optimum. To address these problems and enhance the performance of FA, we propose a new firefly algorithm, which is called the Best Neighbor Firefly Algorithm(BNFA). It employs the best neighbor guided strategy, where each firefly is attracted to the best firefly among some randomly chosen neighbors, thus reducing the firefly oscillations in every attraction-induced migration stage, while increasing the probability of the guidance a new better direction. Moreover, it selects neighbors randomly to prevent the firefly form being trapped into a local optimum. Extensive experiments are conducted to find out the optimal parameter settings. To verify the performance of BNFA, 13 classical benchmark functions are tested. Results show that BNFA outperforms the standard FA and other recently proposed modified FAs. |
