OPPOSITION & DIMENSIONAL BASED FIREFLY ALGORITHM

Abstract

Evolutionary algorithms (EAs) are well-known optimization approaches to deal with nonlinear and complex problems. Evolutionary algorithms often perform well approximating solutions to all types of problems because they ideally do not make any assumption about the underlying fitness landscape. Firefly algorithm is one of the new metaheuristic algorithms for optimization problems. It was originally proposed by X. S. Yang [1]. It is one of the latest additions to the family of swarm intelligence metaheuristics for hard optimization problems. The algorithm is inspired by the flashing behavior of fireflies. In the algorithm, randomly generated solutions will be considered as fireflies, and brightness is assigned depending on their performance on the objective function. One of the rules used to construct the algorithm is, a firefly will be attracted to a brighter firefly, and if there is no brighter firefly, it will move randomly. This thesis presents the modified Firefly Algorithm (FA). The proposed approach gives more efficient solution with reduced time complexity in comparison to original FA. Two modifications made are; 1) Opposition-based methodology is deployed where initialization of candidate solutions is done using opposition based learning to improve convergence rate of original FA. In this opposite numbers are used which helps in generating the initial solutions from both ends hence explores the search space more efficiently. 2) The dimensional-based approach is employed in which optimization of each dimension of the solution is done separately. The dimensional method is specifically employed to conquer the “curse of dimensionality,” by splitting a firefly with composite high dimensionality into several 1-D subparts. Then, the firefly makes contribution to the population not only as a whole item but also in each dimension. This ensures searching the optimal value of each dimension efficiently and hence gives more optimal solution. Several complex multidimensional standard functions are employed for experimental verification. Experimental results show that the ODFA (Opposition and Dimensional based FA) gives more accurate optimal solution with high convergence speed than the original FA