To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Effects of the Number of Constraints on the Performance of Multi-Objective Evolutionary Algorithms |
|
Author |
Yuki Tanigaki, Naoki Masuyama, and Yusuke Nojima |
| Citation |
Vol. 18 No. 12 pp. 221-231 |
|
Abstract |
In recent years, there is a great demand for algorithms solving constrained multi-objective optimization problems (CMOPs) in real-world engineering applications. Unlike the box-constrained problems, an additional constraint handling technique (CHT) is required to solve CMOPs. In real-world engineering applications, there are cases in which 50 or more constraints are considered simultaneously. Thus, it is important to examine the behavior of each CHT on a large number of constraints. However, well-known test problems such as CF functions and C-DTLZ functions have only a small number of constraints. One or two constraints are considered among CF functions and all problems included in C1-DTLZ and C2-DTLZ functions have a single constraint. In this context, we propose test problems that can freely change the number of constraints. In designing such test problems, we extend the WFG toolkit, which is proposed for creating box-constrained multi-objective test problems. The performance of a number of popular CHTs is compared on the proposed test problems with up to 100 constraints. We can observe severe performance deterioration by the increase in the number of constraints. |
|
Keywords |
Benchmark Problem Framework, Constraint Handling, Evolutionary Multi-Objective Optimization, WFG Toolkit. |
|
URL |
