Research map / Probabilistic and scientific modeling / Bayesian optimization
Discovering Evolution Strategies via Meta-Black-Box Optimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the attention-based learned evolution strategy instead of established hand-crafted ES | Neuroevolution policies · Brax continuous control | normalized performance | Helped |
| Use the attention-based learned evolution strategy | CNN classifiers · MNIST variants | test accuracy | Helped |
| Use the attention-based learned evolution strategy | BBOB optimization · BBOB functions | optimization performance | Helped |
| Fix recombination weights to the rank-based scheme (ablation) | LES · Brax continuous control | normalized performance | Hurt |
| Fix learning rates to α_m=1 and α_σ=0.1 (ablation) | LES · Brax continuous control | normalized performance | No clear effect |
| Fix both recombination weights and learning-rate modulation (ablation) | LES · Brax continuous control | normalized performance | Hurt |
| Broaden the meta-training task distribution | LES · Brax continuous control | normalized performance | Helped |
| Increase the meta-training population size alongside a broader task set | LES · Brax continuous control | normalized performance | Helped |
| Replace neural recombination weights with the rank-based softmax approximation at β≈12.5 | LES, yielding DES · Brax continuous control | normalized performance | No clear effect |
| Tune the DES temperature parameter β | DES · Brax continuous control | normalized performance | Helped |
| Use a frozen pretrained LES as the outer-loop meta-optimizer | MetaBBO · Brax continuous control | normalized performance | Hurt |
| Use self-referential meta-training from random initialization | MetaBBO · Brax continuous control | normalized performance | Helped |
Closest papers by meaning
- Guided evolutionary strategies: augmenting random search with surrogate gradients (ICML 2019 · similarity 0.90)
- ConfigX: Modular Configuration for Evolutionary Algorithms via Multitask Reinforcement Learning (AAAI 2025 · similarity 0.87)
- MetaBox-v2: A Unified Benchmark Platform for Meta-Black-Box Optimization (NeurIPS 2025 · similarity 0.87)
- Teacher-Aware Evolution of Heuristic Programs from Learned Optimization Policies (NeurIPS 2026 · similarity 0.87)
- Diffusion Models are Evolutionary Algorithms (ICLR 2025 · similarity 0.86)
- Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.