Evolutionary Computation for Multicomponent Problems: Opportunities and Future Directions

Abstract

Over the past 30 years many researchers in the field of evolutionary computation have put a lot of effort to introduce various approaches for solving hard problems. Most of these problems have been inspired by major industries so that solving them, by providing either optimal or near optimal solution, was of major significance. Indeed, this was a very promising trajectory as advances in these problem-solving approaches could result in adding values to major industries. In this paper we revisit this trajectory to find out whether the attempts that started three decades ago are still aligned with the same goal, as complexities of real-world problems increased significantly. We present some examples of modern real-world problems, discuss why they might be difficult to solve, and whether there is any mismatch between these examples and the problems that are investigated in the evolutionary computation area.1 See (Michalewicz, 2012b) for details on different interpretations of the term "real-world problems".

Multifactorial Evolution: Toward Evolutionary Multitasking

“…However, it is worth highlighting here that the applicability of MFO is not restricted to such scenarios. It can often happen that a single complex problem is composed of several interdependent sub-problems or components [47]- [49]. MFO can conceivably be adapted to solve these components simultaneously.…”

Decomposition for Large-scale Optimization Problems with Overlapping Components

“…Methodology: In this sub-section, we perform three sets of comparisons on the CEC'2013 benchmark problems: 1) RDG3 (with ǫ n = 50 and ǫ s = 100) versus RDG, RDG2, DG2 and GDG; 2) CC versus CBCC (used in Section IV-D), with RDG3 as the decomposition method; and 4) CC-RDG3 versus 9 stateof-the-arts listed in the TACO website. 1 Results: We observe in Table III that RDG3 significantly outperforms the other four decomposition methods on overlapping problems f 13 and f 14 . RDG3 can generate significantly better solution quality than RDG2 for problems with separable variables e.g., f 1 , f 2 and f 4 , suggesting it is useful to further decompose separable variables into small components.…”

“…Many real-world large-scale optimization problems consist of several small sub-problems (or components) that possibly interact with each other [1]- [4]. Exploiting module structure can greatly facilitate solving such a problem [5], [6].…”

A case study of algorithm selection for the traveling thief problem

“…One of these reasons is that academic experiments focused on single component (single silo) benchmark problems, whereas real-world problems are often multi-component problems. In order to guide the community towards this increasingly important aspect of real-world optimization [8], the traveling thief problem was introduced [6] in order to illustrate the complexities that arise by having multiple interacting components.…”

“…These two components have been merged in such a way that the optimal solution for each single one does not necessarily correspond to an optimal TTP solution. The motivation for the TTP is to allow the systematic investigation of interactions between two hard component problems, to gain insights that eventually help solve real-world problems more efficiently [8].…”