Firefly Algorithm (FA) is one of the new natural inspired optimization algorithms. It is inspired by the flashing behavior of the fireflies. Firefly algorithm, has some drawbacks such as getting trapped into several local optima, FA parameters are set fixed without change during iterations time. Besides that, it does not memorize or remember the history of any situation for each iteration. In this paper, we propos a firefly photinus algorithm (FPA) based on the initialize mate list to solve problems of trapped into several local optima and remember history of situation to forbidden fireflies movements in mate list (history) during the search process, and propose new absorption parameter r to change the parameters during iterations time which lead to balance between exploration and exploitation, and it controls the dominance area of a lighter firefly during time iterations by reduction or increase r coefficient whether. The experimental results tested on thirteen benchmark functions are selected to evaluate performance of the FPA and to compare it with the standards of the FA and Some FA variants algorithm, it show that FPA algorithm can outperform FA and FA variants algorithm in most of the experiments. (C) 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University.

The artificial bee colony (ABC) algorithm is a relatively new algorithm inspired by nature and has been shown to be efficient in contrast to other optimization algorithms. Nonetheless, ABC has some similar drawbacks to the optimization algorithms in terms of the unbalanced search behavior. The original ABC algorithm shows strong exploration capability with ineffective exploitation due to the unbalanced search model. In this paper, a new ABC algorithm called MeanABC is introduced to achieve the search behavior balance via a modified search equation based on the information of the mean of the previous best solutions. To evaluate the performance of the proposed algorithm, experiments were divided into two parts: First, the proposed algorithm was tested on a comprehensive set of 14 benchmark functions. The results show that the proposed MeanABC enhances the performance of the original ABC in terms of faster global convergence speed, solution quality, and better robustness when compared to other ABC variants. Secondly, the proposed algorithm was applied as a hybrid with the FCM algorithm as a segmentation technique to a set of 20 volumes of real brain MRI images with 20 images for each volume. All of these images have several characteristics, levels of difficulty, and cover different domains. The obtained results are promising, especially when the performance of the proposed algorithm was compared to other state-of-the-art segmentation techniques.