1559. Exploring the Impact of Objective Correlations in Multi-objective Optimisation
Invited abstract in session WB-51: Machine learning approaches in multiobjective decision making, stream Multiobjective and vector optimization.
Wednesday, 10:30-12:00Room: Parkinson B22
Authors (first author is the speaker)
| 1. | Tinkle Chugh
|
| Computer Science, University of Exeter | |
| 2. | Samuel Tebbet
|
| University of Exeter | |
| 3. | George De Ath
|
Abstract
Multi-objective optimisation problems (MOPs) typically involve multiple, often conflicting objectives which can often be computationally expensive to evaluate. Multi-objective optimisers such as multi-objective evolutionary algorithms (MOEAs) provide a computationally feasible way of solving MOPs. Despite the availability of numerous continuous benchmark problems, there is a scarcity of multi-objective optimisation problems that consider the impact of interdependencies or correlations between objectives. This study explores the influence of these correlations on bi-objective optimisation problems in continuous search space. We investigate the effect of correlations from perfect negative to perfect positive on the performance of multi-objective optimisers, specifically MOEAs. Experiments with two objectives and 2--20 decision variables across three MOEA categories (dominance-, indicator-, and decomposition-based) demonstrate the effects of varying correlations. The analysis and results can provide insights in selecting an optimiser for solving similar real-world problems.
Keywords
- Global Optimization
- Continuous Optimization
- Machine Learning
Status: accepted
Back to the list of papers