Research map / Optimization / Optimizers
NeuroEvoBench: Benchmarking Evolutionary Optimizers for Deep Learning Applications
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 a larger candidate population instead of increasing evaluations per candidate (ablation) | OpenAI-ES · Noisy Ant control | episode return | Helped |
| Increase population size (ablation) | OpenAI-ES · Ant, Fetch, Asterix, Breakout, Fashion-MNIST, CIFAR-10, Addition, and S-MNIST | task performance | Helped |
| Increase model capacity (ablation) | OpenAI-ES · Control, vision, and sequence tasks | task performance | Mixed |
| Use small mean multiplicative decay (ablation) | OpenAI-ES · Neuroevolution tasks | test performance | Helped |
| Use too much mean multiplicative decay (ablation) | OpenAI-ES · Neuroevolution tasks | test performance | Hurt |
| Use centered-rank fitness transformation as the default (ablation) | OpenAI-ES · Neuroevolution tasks | task performance | Helped |
| Use raw-return fitness instead of centered ranks (ablation) | OpenAI-ES · Fetch control | task performance | Helped |
| Change the gradient optimizer (ablation) | OpenAI-ES · Three NEB tasks | task performance | No clear effect |
| Choose Adan as the gradient optimizer (ablation) | OpenAI-ES · Fetch control | task performance | Helped |
| Use a Genetic Algorithm instead of an Evolution Strategy | Neuroevolution optimizers · Nine neuroevolution tasks | normalized median performance | Hurt |
| Use a meta-learned optimizer rather than a manually designed optimizer | LES and LGA · Ant and Fetch control | task performance | Helped |
| Apply meta-learned optimizers beyond their meta-training setting | LES and LGA · Neuroevolution tasks | task performance | Hurt |
Closest papers by meaning
- EVOS: Efficient Implicit Neural Training via EVOlutionary Selector (CVPR 2025 · similarity 0.87)
- Runtime Analysis of Evolutionary NAS for Multiclass Classification (ICML 2025 · similarity 0.86)
- Evolution Strategies at the Hyperscale (ICML 2026 · similarity 0.86)
- Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks (NeurIPS 2026 · similarity 0.85)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.85)
- Accelerating neural network training: An analysis of the AlgoPerf competition (ICLR 2025 · similarity 0.84)
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.