Research map / Optimization / Optimizers
On the Convergence of Prior-Guided Zeroth-Order Optimization Algorithms
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 |
|---|---|---|---|
| add historical prior to RGF at lr 0.2 (ablation) | RGF · MNIST attack | median queries: 777 -> 484 | Helped |
| add historical prior to RGF at lr 0.1 (ablation) | RGF · MNIST attack | median queries: 1596 -> 572 | Helped |
| add historical prior to ARS at lr 0.2 (ablation) | ARS · MNIST attack | median queries: 735 -> 484 | Helped |
| add historical prior to ARS at lr 0.1 (ablation) | ARS · MNIST attack | median queries: 1386 -> 550 | Helped |
| halve learning rate from 0.2 to 0.1 (ablation) | RGF · MNIST attack | median queries: 777 -> 1596 | Hurt |
| halve learning rate from 0.2 to 0.1 (ablation) | History-PRGF · MNIST attack | median queries: 484 -> 572 | Hurt |
| replace RGF with ARS (accelerated random search) without prior (ablation) | RGF · MNIST attack | median queries: 777 -> 735 | No clear effect |
| replace History-PRGF with History-PARS at lr 0.05 (ablation) | History-PRGF · MNIST attack | median queries: 704 -> 726 | No clear effect |
| use prior-guided gradient estimators instead of RGF/ARS (ablation) | RGF/ARS · numerical benchmarks f1-f3, d=256 | log relative suboptimality | Helped |
| use PARS instead of PARS-Naive (PRGF estimator swapped into ARS) (ablation) | ARS · numerical benchmarks f1-f3 | log relative suboptimality | Helped |
| use ARS-based methods instead of greedy descent (ablation) | greedy descent (RGF/PRGF) · ill-conditioned benchmarks f1-f3 | log relative suboptimality | Helped |
| use historical prior with learning rate 0.02 of optimal (ablation) | RGF/ARS · f1, f2, d=500 | convergence rate | Helped |
Closest papers by meaning
- Refining Adaptive Zeroth-Order Optimization at Ease (ICML 2025 · similarity 0.90)
- Robust and Faster Zeroth-Order Minimax Optimization: Complexity and Applications (NeurIPS 2024 · similarity 0.89)
- DeepZero: Scaling Up Zeroth-Order Optimization for Deep Model Training (ICLR 2024 · similarity 0.87)
- Adaptively Incorporating Directional Hints into Zeroth-Order Optimization (NeurIPS 2026 · similarity 0.87)
- On the Optimal Construction of Unbiased Gradient Estimators for Zeroth-Order Optimization (NeurIPS 2025 · similarity 0.87)
- Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles (ICLR 2024 · similarity 0.87)
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.