Research map / Probabilistic and scientific modeling / Molecules and proteins
Efficient Non-Parametric Optimizer Search for Diverse Tasks
Ruochen Wang, Yuanhao Xiong, Minhao Cheng, Cho-Jui Hsieh
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 |
|---|---|---|---|
| replace random search with MCT traversal (ablation) | MCT optimizer search · MNISTNET | test accuracy: 88.02% -> 89.43% | Improved |
| replace random search with MCT traversal (ablation) | MCT optimizer search · MNISTNET | training loss (sum): 60.25 -> 59.25 | Improved |
| disable score thresholding (ablation) | MCT optimizer search · MNISTNET | cumulative training loss: 59.25 -> 60.25 | Hurt |
| search with MCT on super-tree instead of RL on hand-crafted patterns | Optimizer search · CIFAR-10 ConvNet | test accuracy: 76.21% -> 77.02% | Improved |
| use 128 evaluations instead of over 10,000 | Optimizer search vs NOS-RL · CIFAR-10 ConvNet | search budget (#evaluations): >10,000 -> 128 | Improved |
| use discovered optimizer instead of Adam | GAT · OGBN-Products | accuracy: 77.49% -> 80.15% | Improved |
| use discovered optimizer instead of Adam | GAT · Citeseer | accuracy: 71.70% -> 73.10% | Improved |
| use discovered optimizer instead of AdamW | BERT-base · MRPC | accuracy: 82.84 -> 86.64 | Improved |
| use discovered optimizer instead of AdamW | BERT-base · RTE | accuracy: 65.97 -> 68.50 | Improved |
| use log-based discovered update instead of sign-based PGD | PGD attack · CIFAR-10 Carmon2019 | attack success rate: 37.83% -> 38.35% | Improved |
| compare discovered optimizer with Adaptive PGD | APGD · CIFAR-10 Gowal2020 (WRN-70-16) | attack success rate: 32.00% -> 32.00% | No clear effect |
| transfer a single discovered optimizer (products2) across graph datasets | Discovered optimizer vs Adam · PubMed | accuracy: 78.20% -> 77.12% | Hurt |
Closest papers by meaning
- Trace is the Next AutoDiff: Generative Optimization with Rich Feedback, Execution Traces, and LLMs (NeurIPS 2024 · similarity 0.86)
- MM-OptBench: A Solver-Grounded Benchmark for Multimodal Optimization Modeling (NeurIPS 2026 · similarity 0.86)
- Eco Search: A No-delay Best-First Search Algorithm for Program Synthesis (AAAI 2025 · similarity 0.86)
- SolverLLM: Leveraging Test-Time Scaling for Optimization Problem via LLM-Guided Search (NeurIPS 2025 · similarity 0.86)
- OptiTree: Hierarchical Thoughts Generation with Tree Search for LLM Optimization Modeling (NeurIPS 2025 · similarity 0.86)
- Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces (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.
