Multi-objective gene-pool optimal mixing evolutionary algorithms

Abstract

The recently introduced Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA), with a lean, but sufficient, linkage model and an efficient variation operator, has been shown to be a robust and efficient methodology for solving single objective (SO) optimization problems with superior performance compared to classic genetic algorithms (GAs) and estimation-of-distribution algorithms (EDAs). In this paper, we bring the strengths of GOMEAs to the multiobjective (MO) optimization realm. To this end, we modify the linkage learning procedure and the variation operator of GOMEAs to better suit the need of finding the whole Paretooptimal front rather than a single best solution. Based on state-of-the-art studies on MOEAs, we further pinpoint and incorporate two other essential components for a scalable MO optimizer. First, the use of an elitist archive is beneficial for keeping track of non-dominated solutions when the main population size is limited. Second, clustering can be crucial if different parts of the Pareto-optimal front need to be handled differently. By combining these elements, we construct a multi-objective GOMEA (MO-GOMEA). Experimental results on various MO optimization problems confirm the capability and scalability of our MO-GOMEA that compare favorably with those of the well-known GA NSGA-II and the more recently introduced EDA mohBOA.

A Multi-Objective Optimization Framework for Offshore Wind Farm Layouts and Electric Infrastructures

“…The MOGOMEA, recently introduced in the literature [122], is a metaheuristic discrete MO optimization algorithm which demonstrated to outperform, i.e., obtain better optimized trade-offs, other state-of-the-art well-known MO optimization algorithms on standard combinatorial benchmark functions [122], as well as in real-world applications [33,123] for a given budget of function evaluations. Although the MOGOMEA is capable of obtaining high-quality results, it does not guarantee global optimal solutions are found given a finite amount of evaluations of the optimization functions [122].…”

“…The algorithm then alters solutions into offspring by exchanging variables between different solutions. The MOGOMEA clusters solutions that are in the same objective space vicinity and only performs variation within each cluster since solutions tend to be very dissimilar for different areas of the objective space [122].…”

“…In this way, the population size is incrementally increased throughout the optimization [33]. This is done to allow new genetic material into the population since the algorithm performs a very intensive local-search strategy and hence, it may fail to layouts that cannot be reached from the initial solutions (only through mutation) [122].…”

“…Although the MOGOMEA was initially designed to have a population of solutions that are sequentially altered to generate offsprings [122], the version of MOGOMEA employed evaluates the solutions in an asynchronous parallel manner. Currently, it is becoming even more important to move towards a parallel environment in which several CPU cores may be used simultaneously [129].…”

“…The algorithm groups the wind farm layouts into five clusters during the optimization process since this setting has proved to already provide improved results over using a single cluster in recent literature [122].…”