A novel grid-oriented dynamic weight parameter based improved variant of Jaya algorithm

Abstract

Jaya algorithm (JA) is a single-step metaheuristic optimization technique that is free from algorithm-specific parameters. Regardless of its simplicity, JA proved its effective performance against the variety of optimization algorithms (Du et al., 2018). However, like other swarm-based optimization techniques, the JA also suffers from the inadequacies of slow or premature convergence (Farah and Belazi, 2018). In this study, an improved variant of JA (IJaya) is proposed whose functioning depends on the randomly initiated bounds based grid-oriented weight parameters. Initially, aiming to balance the global exploration and local exploitation capabilities of JA, a dynamic weight parameter is introduced as a varying coefficient for the entire solution updating expression of JA. Then, to maintain the population diversity and to mitigate the complexity of parameter tuning, the introduced weight parameter is dealt with the randomly selected parameter bounds based grid-search mechanism. The proposed IJaya algorithm is benchmarked on well-known 15 unconstrained mathematical test functions, and its performance is analyzed against the standard JA, one modified variant of JA, some well-known state-of-the-art, and few newly introduced optimization algorithms. Furthermore, the non-parametric Friedman and Quade rank tests are also conducted which confirmed the superiority of proposed IJaya both in convergence rate and solution quality. The paper also presents the results obtained by IJaya in two classical structural design problems (a cantilever beam and a 3-bar truss) and a real-world electrical power engineering problem. Numerical results clearly prove the efficiency of the proposed algorithm.