Reproduce “CMA-ES with Margin: Lower-Bounding Marginal Probability for Mixed-Integer Black-Box Optimization” 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
Ryoki Hamano, Shota Saito, Masahiro Nomura, Shinichi Shirakawa
Researchers from Yokohama National University and industry partners developed "CMA-ES with Margin," a modified Covariance Matrix Adaptation Evolution Strategy that effectively tackles mixed-integer black-box optimization problems by preventing premature stagnation of discrete variables. The method achieved a 100% success rate on benchmark problems, outperforming existing adaptations, particularly for binary and general integer variables.
Key insights
- Preventing the marginal probabilities of discrete variable values from dropping below a specified margin (alpha) is critical to sustain exploration and avoid premature convergence in MI-BBO.
- Decoupling the handling of integer variables through a post-processing step on the mean and an adaptive scaling matrix allows CMA-ES to retain its inherent strengths, like affine invariance, while robustly optimizing mixed-integer problems.
- The margin parameter alpha, empirically set to (N λ)^-1 (where N is dimension and λ is population size), offers a reliable balance between maintaining sufficient exploration and enabling convergence for diverse problem types.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.