Effect of Population Size Over Parameter-less Firefly Algorithm

Nature-inspired optimization algorithms have proved their efficacy for solving highly nonlinear problems in science and engineering. Firefly Algorithm (FA), which is one of them, has been taken to be broadly discussed here. However, practical studies demonstrate that the proper setting of the FA’s parameters is the key difficulty. FA associates with some sensitive parameters, and proper setting of their values is extremely time-consuming. Therefore, parameter less variants of FA have been proposed in this study by incorporating the adaptive formulation of the control parameters which are free from the tuning procedure. Population size greatly influences the convergence of any nature-inspired optimization algorithm. Generally, lower population size enhances convergence speed but may trap into local optima. While larger population size maintains the diversity but slows down the convergence rate. Therefore, the crucial effect of different population size over FA’s efficiency has also been investigated here. The population size (of considered FA is varied within the size [10, 1280] where it gets doubled in each run starting with \(n = 10\). Therefore, eight instances of FA have been executed, and the best one is considered by the user over different classes of functions. Experimental results also show that the proposed parameter-less FA with a population size between 40 and 80 provides best results by considering optimization ability and consistency over any classes of functions.