Reproduce “A Non-Dominated Sorting Evolutionary Algorithm Updating When Required” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lucas R. C. Farias, Abimael J. F. Santos, Matheus R. B. Nobre
A hybrid many-objective evolutionary algorithm, NSGA-III-UR, dynamically assesses Pareto optimal front (POF) regularity to conditionally activate adaptive reference vector mechanisms. This approach allowed it to achieve competitive or superior performance compared to NSGA-III and A-NSGA-III across diverse benchmark and real-world optimization problems, with 15 out of 27 best IGD values and consistently better performance on real-world cases.
Key insights
- Dynamically assessing Pareto optimal front (POF) regularity via a metric like the Spreading Index enables an algorithm to intelligently switch between static and adaptive reference vector strategies.
- Conditional adaptation, triggered by an estimated POF irregularity, can reconcile the efficiency of fixed reference points for regular problems with the necessity of adaptive schemes for irregular ones, optimizing computational resource use.
- Integrating a statistical threshold based on the number of objectives provides a robust decision-making process for adaptation, preventing frequent mode switching while maintaining structural integrity for regular POFs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.