Research map / Optimization / Nonconvex and smooth optimization
Scaling Simultaneous Optimistic Optimization for High-Dimensional Non-Convex Functions with Low Effective Dimensions
Hong Qian, Yang Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| underestimate effective dimension (d=1 < de=2) (ablation) | RESOO · 1000-d Branin | simple regret: 0.004 -> 3.816 | Hurt |
| use smaller estimated upper bound d (d=2 instead of d=4) at M=2 (ablation) | RESOO · 1000-d Branin | simple regret: 0.075 -> 0.002 | Improved |
| increase restarts M from 1 to 2 (d=2) (ablation) | RESOO · 1000-d Branin | simple regret: 0.004 -> 0.002 | Improved |
| increase restarts M from 4 to 10 (d=2) (ablation) | RESOO · 1000-d Branin | simple regret: 0.001 -> 0.107 | Hurt |
| increase restarts M from 4 to 10 (d=4) (ablation) | RESOO · 1000-d Rosenbrock | simple regret: 1.556 -> 2.691 | Hurt |
| random embedding with restarts (RESOO) instead of SOO | SOO · rotated Branin and Rosenbrock, D=100 to 100000 | simple regret | Improved |
| increase ambient dimension D from 100 to 100000 | RESOO · Branin and Rosenbrock | simple regret | No clear effect |
| use RESOO instead of SOO with separated hyper-parameters | SOO · MNIST SVM tuning | test accuracy: 94.33% -> 94.81% | Improved |
| use RESOO instead of SOO with separated hyper-parameters | SOO · Letter SVM tuning | test accuracy: 85.06% -> 85.47% | Improved |
| separate per-pair SVM hyper-parameters instead of one shared value (random search) | multi-class SVM · MNIST | test accuracy: 93.75% -> 94.13% | Improved |
| separate per-pair SVM hyper-parameters instead of shared value (SOO) | SOO · MNIST | test accuracy: 94.61% -> 94.33% | Hurt |
| larger effective-dimension estimate d2 vs d1 (ablation) | RESOO · Vowel SVM tuning | test accuracy: 83.40% -> 83.16% | Hurt |
Closest papers by meaning
- Solving High-Dimensional Multi-Objective Optimization Problems with Low Effective Dimensions (AAAI 2017 · similarity 0.87)
- Adaptive Partitioning Schemes for Optimistic Optimization (ICML 2025 · similarity 0.85)
- Adaptive and Optimal Second-order Optimistic Methods for Minimax Optimization (NeurIPS 2024 · similarity 0.84)
- Double Variance Reduction: A Smoothing Trick for Composite Optimization Problems without First-Order Gradient (ICML 2024 · similarity 0.83)
- Distributionally Robust Optimization with Bias and Variance Reduction (ICLR 2024 · similarity 0.83)
- Drago: Primal-Dual Coupled Variance Reduction for Faster Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.83)
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.
