Runtime Analysis of Cross-Party Recombination in Multi-Party Multi-Objective Optimization
The research documented in arXiv (2605.17454) explores the runtime performance of cross-party recombination in multi-party multi-objective optimization problems (MPMOPs), which necessitate agreement among independent decision-makers. The authors demonstrate that a mutation baseline guided by payoff encounters a gap-crossing limitation, demanding Θ(n²) expected fitness evaluations on the MP-JCG benchmark. In contrast, a variant of CPR-NSGA-II effectively identifies common Pareto-optimal solutions with O(n log n) expected evaluations by utilizing complementary prefix and suffix templates from various party populations. This study emphasizes the superior efficiency of cross-party recombination compared to traditional many-objective approaches.
Key facts
- arXiv paper 2605.17454 analyzes runtime of cross-party recombination for MPMOPs
- MPMOPs require consensus among autonomous decision makers
- Payoff-guided mutation baseline requires Θ(n²) expected evaluations on MP-JCG
- CPR-NSGA-II variant achieves O(n log n) expected evaluations
- Cross-party recombination assembles complementary prefix and suffix templates
- Study compares with flattened four-objective formulation F-JCG
- Existing runtime theory is tailored to single-party Pareto-front approximation
- MP-JCG is a pseudo-Boolean benchmark with an explicit gap region
Entities
Institutions
- arXiv