An improved firefly algorithm for numerical optimisation

Firefly algorithm (FA) is a recently proposed meta-heuristic optimisation technique, which has shown good performance on many optimisation problems. In the original FA, each firefly is attracted by any other brighter firefly (better fitness value). By the attraction, fireflies maybe moved to better positions. However, the attraction does not guarantee whether a firefly is moved to a better position. Sometimes, the attraction may move a firefly to a worse position. Therefore, the search of firefly is oscillated during the evolution. In this paper, we present an improved firefly algorithm (IFA), which employs a greedy selection method to guarantee that a firefly is not moved to worse positions. To verify the performance of IFA, a set of well-known benchmark functions are used in the experiments. Experimental results show that the IFA achieves better results than the original FA.