Diversity-adaptive parallel memetic algorithm for solving large scale combinatorial optimization problems

Abstract

Parallel memetic algorithms (PMAs) are a class of modern parallel meta-heuristics that combine evolutionary algorithms, local search, parallel and distributed computing technologies for global optimization. Recent studies on PMAs for large-scale complex combinatorial optimization problems have shown that they converge to high quality solutions significantly faster than canonical GAs and MAs. However, the use of local learning for every individual throughout the PMA search can be a very computationally intensive and inefficient process. This paper presents a study on two diversity-adaptive strategies, i.e., (1) diversity-based static adaptive strategy (PMA-SLS) and (2) diversitybased dynamic adaptive strategy (PMA-DLS) for controlling the local search frequency in the PMA search. Empirical study on a class of NP-hard combinatorial optimization problem, particularly large-scale quadratic assignment problems (QAPs) shows that the diversityadaptive PMA converges to competitive solutions at significantly lower computational cost when compared to the canonical MA and PMA. Furthermore, it is found that the diversity-based dynamic adaptation strategy displays better robustness in terms of solution quality

A case study of memetic algorithms for constraint optimization

“…MAs aim to balance the exploration and exploitation capabilities of both genetic algorithms and local search. Many researchers already highlighted the effectiveness of integrating meme(s) into evolutionary algorithms for solving complex optimization problems based on various frameworks [1,2,13,30,38,40,48,53,54]. In a canonical MA, a prefixed single meme is employed after mutation and evaluation steps of a GA. Obviously, a variety of memes might be designed for solving a specific problem.…”

Meta-Lamarckian learning in three stage optimal memetic exploration

“…In [4] the classification of coordination methods has been extended and updated. The following four categories have been identified: 1) Adaptive Hyper-heuristic, where heuristic rules are employed (e.g., [8], [9], [10], [11]); 2) MetaLamarckian learning defined in [12], where the activation of the memes depends on their success, see also [13], [14], [15]; 3) Self-Adaptive and Co-Evolutionary, where the rules coordinating the memes are evolving in parallel with the candidate solutions of the optimization problem or encoded within the solution, see [16], [17], [18], [19]; and 4) Fitness DiversityAdaptive, where the activation of the memes depends on a measure of the diversity (e.g., [20], [21], [6], [22], [23]). As a general idea, the algorithmic designer attempts to have a system which performs the coordination automatically.…”

Evolutionary-based Image Segmentation Methods

“…Recently, MAs have been well used across a wide range of problem domains. A lot of studies have demonstrated that MAs converge to high-quality solutions more efficiently than their conventional counterparts in many realworld applications [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32]. MAs have also been applied in the image processing field [33][34][35][36][37][38].…”