Recover Fractional Lévy Flight Bat Algorithm for Optimization in Artificial Neural Networks Model

Abstract

The bat algorithm (BA) is a heuristic optimization method based on the loudness and the rate of ultrasonic bursts. It has been simulated bats use echolocation for foraging. It has been proven that this algorithm has a good ability to search for the global optimum, but it suffers from slow searching speed in the last iterations. This paper proposes a recover fractional Lévy flight bat algorithm (rFLFBA) to improve BA by adjusting the inertia weight of velocity update instead of fractional Lévy flight from fractional Lévy flight bat algorithm (FLFBA). rFLFBA is deployed as learning method for artificial neural networks (ANNs) to increase the efficiencies in avoiding the local minima problem and the slow convergence rate of bat algorithms. The results are compared with a Particle Swarm optimization and Gravitational Search Algorithm (PSOGSA), Grey Wolf optimizer (GWO), and FLFBA learning methods for ANNs. The resulting classification rate of ANNs trained with GWO, PSOGSA, and FLFBA is also examined. The simulation results show that rFLFBA better than GWO and PSOGSA for training ANNs in terms of convergence curve. It is also proven that an ANN trained with rFLFBA has better accuracy than one trained with FLFBA.