[Paper Review] Bat Algorithm is Better Than Intermittent Search Strategy

Xin‐She Yang, Suash Deb+1 more|arXiv (Cornell University)|Aug 22, 2014

Diffusion and Search Dynamics18 references15 citations

TL;DR

This paper proposes that the bat algorithm (BA) outperforms the intermittent search strategy (ISS) in balancing local exploitation and global exploration for global optimization. Through simulations on benchmark functions and real-world engineering problems, BA achieves superior convergence speed and solution quality with fewer function evaluations, demonstrating its efficiency in high-dimensional and constrained optimization tasks.

ABSTRACT

The efficiency of any metaheuristic algorithm largely depends on the way of balancing local intensive exploitation and global diverse exploration. Studies show that bat algorithm can provide a good balance between these two key components with superior efficiency. In this paper, we first review some commonly used metaheuristic algorithms, and then compare the performance of bat algorithm with the so-called intermittent search strategy. From simulations, we found that bat algorithm is better than the optimal intermittent search strategy. We also analyse the comparison results and their implications for higher dimensional optimization problems. In addition, we also apply bat algorithm in solving business optimization and engineering design problems.

Motivation & Objective

• To investigate the balance between local exploitation and global exploration in metaheuristic algorithms.
• To evaluate the performance of the bat algorithm against the intermittent search strategy (ISS), a theoretical model of search behavior.
• To determine whether BA can achieve better optimization efficiency with fewer function evaluations than ISS.
• To validate BA’s effectiveness on real-world engineering and business optimization problems.
• To provide theoretical and empirical insights into algorithm-dependent parameter tuning for improved performance.

Proposed method

• The bat algorithm is applied to 2D multimodal objective functions to analyze its exploitation-exploration dynamics.
• The intermittent search strategy (ISS) is used as a benchmark for comparison, based on theoretical models of random walk-based search with periodic pauses.
• Simulations are conducted across multiple test functions to compare convergence speed and solution quality between BA and ISS.
• The bat algorithm uses frequency tuning, velocity update, and local search mechanisms inspired by echolocation in bats.
• The algorithm is tested on constrained engineering problems, including spring design and welded beam design, with explicit objective functions and constraints.
• Support vector machine (SVM) benchmarking is used to validate performance on a business optimization problem with 30 activities.

Experimental results

Research questions

• RQ1 Can the bat algorithm achieve a better balance between local exploitation and global exploration than the intermittent search strategy?
• RQ2 How does the bat algorithm compare to ISS in terms of convergence speed and solution accuracy on multimodal test functions?
• RQ3 What is the performance of the bat algorithm on high-dimensional and constrained engineering optimization problems?
• RQ4 Can the bat algorithm reduce the number of function evaluations required to reach optimal or near-optimal solutions compared to ISS?
• RQ5 What are the implications of the observed differences between theoretical ISS models and actual metaheuristic behavior for algorithm design?

Key findings

• The bat algorithm achieved a mean deviation of 0.049 from the optimal solution on the J=30 job shop scheduling problem with 5000 iterations, outperforming PSO (0.21) and hybrid GA (0.06).
• In the spring design problem, the bat algorithm found a solution of f* = 0.012665 at (0.051690, 0.356750, 11.287126), demonstrating high accuracy and constraint satisfaction.
• For the welded beam design, the bat algorithm achieved a minimum cost of f(x*) = 1.724852308598 at the optimal design vector (0.2057, 3.4705, 9.0366, 0.2057).
• The bat algorithm required significantly fewer function evaluations than the intermittent search strategy to reach comparable or better solutions.
• The study found a substantial gap between theoretical intermittent search models and the actual behavior of metaheuristics like BA, indicating limitations in current theoretical frameworks.
• Empirical results confirm that the bat algorithm provides a superior balance of exploration and exploitation, especially in high-dimensional and nonlinear optimization problems.