Reproduce “Cascading CMA-ES Instances for Generating Input-diverse Solution Batches” 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
Maria Laura Santoni, Christoph Dürr, Carola Doerr, Mike Preuss, Elena Raponi
CMA-ES-Diversity Search (CMA-ES-DS) is presented as an optimization algorithm capable of generating batches of high-quality, input-space diverse solutions by cascading multiple CMA-ES instances with dynamic tabu regions. The method reliably produces full batches meeting explicit distance constraints while maintaining competitive solution quality, outperforming other approaches especially under stringent diversity requirements.
Key insights
- Enforcing input-space diversity through a cascading system of tabu regions within multiple CMA-ES instances enables the reliable generation of high-quality, diverse solution batches.
- The choice of subset selection method significantly impacts computational efficiency; the "Clearing Process" offers a practical balance between speed and performance for extracting diverse batches from a generated portfolio.
- CMA-ES-DS's explicit strategy for promoting diversity allows it to overcome limitations of traditional global and multimodal optimizers when faced with strict input-space distance constraints, making it robust across various function types and search budgets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.