Back to the roots : multi-X evolutionary computation
Over the years, evolutionary computation has come to be recognized as one of the leading algorithmic paradigms in the arena of global black box optimization. The distinguishing facets of evolutionary methods, inspired by Darwin’s foundational principles of natural selection, stem mainly from their p...
Saved in:
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/143183 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Over the years, evolutionary computation has come to be recognized as one of the leading algorithmic paradigms in the arena of global black box optimization. The distinguishing facets of evolutionary methods, inspired by Darwin’s foundational principles of natural selection, stem mainly from their population-based search strategy—which gives rise to the phenomenon of implicit parallelism. Precisely, even as an evolutionary algorithm manipulates a population of a few candidate solutions (or: individuals), it is able to simultaneously sample, evaluate, and process a vast number of regions of the search space. This behavior is in effect analogous to our inherent cognitive ability of processing diverse information streams (such as sight and sound) with apparent simultaneity in different regions of our brain. For this reason, evolutionary algorithms have emerged as the method of choice for those search and optimization problems where a collection of multiple target solutions (that may be scattered throughout the search space) are to be found in a single run. With the above in mind, in this paper we return to the roots of evolutionary computation, with the aim of shedding light on a variety of problem settings that are uniquely suited for exploiting the implicit parallelism of evolutionary algorithms. Our discussions cover established concepts of multi-objective and multi-modal optimization, as well as new (schema) theories pertaining to emerging problem formulations that entail multiple searches to be carried out at once. We capture associated research activities under the umbrella term of multi-X evolutionary computation, where X, as of now, represents the following list: {“objective,” “modal,” “task,” “level,” “hard,” “disciplinary,” “form”}. With this, we hope that the present position paper will serve as a catalyst for effecting further research efforts into such areas of optimization problem-solving that are well-aligned with the fundamentals of evolutionary computation; in turn prompting the steady update of the list X with new applications in the future. |
|---|