Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The Shift Toward Open and Reproducible AI Research," Under Review, 2026. doi: 10.48550/arXiv.2606.16974..
Convergence Rates of Constrained Expected Improvement
Authors: Haowei Wang, Jingyi Wang, Zhongxiang Dai, Nai-Yuan Chiang, Szu Hui Ng, Cosmin Petra
NeurIPS 2025 | Venue PDF | LLM Run Details | Input Tokens: 27,550 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 3,509 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments are performed to validate the theoretical analysis. |
| Researcher Affiliation | Collaboration | Haowei Wang National University of Singapore Singapore EMAIL Jingyi Wang Lawrence Livermore National Laboratory Livermore, CA 94550 EMAIL Nai-Yuan Chiang Lawrence Livermore National Laboratory Livermore, CA 94550 EMAIL Zhongxiang Dai The Chinese University of Hong Kong, Shenzhen China EMAIL Szu Hui Ng National University of Singapore Singapore EMAIL Cosmin G. Petra Lawrence Livermore National Laboratory Livermore, CA 94550 EMAIL |
| Pseudocode | Yes | Algorithm 1 CEI algorithm 1: Choose kf( , ), kc( , ), and T0 initial samples xi, i = 1, . . . , T0. Observe f1:T0 and c1:T0. |
| Open Source Code | Yes | Codes are available in https://github.com/Haowei-Wang/Convergence-Rates-of-Constrained-Expected Improvement. |
| Open Datasets | Yes | we conduct numerical experiments to support the theoretical results. We apply the CEI algorithm to eight synthetic problems that are randomly generated from RKHS of kernels and GP priors, and five benchmark problems commonly used in the CBO literature. |
| Dataset Splits | Yes | The number of initial design is set to 10d, and 50 optimization iterations were performed for all cases. |
| Hardware Specification | Yes | All experiments are conducted on M1 (16GB memory) |
| Software Dependencies | No | No specific software versions for dependencies are explicitly mentioned in the paper. |
| Experiment Setup | Yes | For each synthetic problem, we conducted 100 independent trials. The number of initial design is set to 10d, and 50 optimization iterations were performed for all cases. ... The SE kernel is used for the GP modeling (similar performance is observed for the Matérn kernel) and the hyper-parameters are estimated by a standard maximum likelihood method. |