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 embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
BayeSQP: Bayesian Optimization through Sequential Quadratic Programming
Authors: Paul Brunzema, Sebastian Trimpe
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical results show that Baye SQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization. |
| Researcher Affiliation | Academia | Paul Brunzema Sebastian Trimpe Institute for Data Science in Mechanical Engineering RWTH Aachen University Aachen, Germany EMAIL |
| Pseudocode | No | The paper describes the algorithm Baye SQP through textual descriptions and a flowchart (Figure 1: Overview) but does not contain a formal 'Pseudocode' or 'Algorithm' block with structured steps. |
| Open Source Code | Yes | The implementation is provided under https://github.com/brunzema/ bayesqp and easily accessible via Py PI. |
| Open Datasets | Yes | Lastly, we also evaluate Baye SQP on standard BO benchmarks. Here, we follow recent best practices and initialize lengthscales with d for all baselines [30, 67, 48]. The results are summarized in Table 1. We can clearly observe that Baye SQP is sensitive to initialization highlighted by the large 90% quantile especially for Ackley. This is to be expected as the algorithm is local and Ackley is very multi-modal. Still, importantly, Baye SQP is able to find feasible solutions for all seeds in all benchmarks contrary to C-log EI. To demonstrate the real-world applicability of Baye SQP, we compare constrained optimization baselines on the 7-dimensional Speed Reducer benchmark [36] For the constrained Gramacy benchmark, we use the formulation introduced in Gramacy et al. [20] |
| Dataset Splits | No | In all subsequent plots, we present the median alongside the 5th to 95th percentile range (90% inner quantiles) computed across 32 independent random seeds. For the within-model comparisons, we freeze the lengthscales of the kernel and do not perform any hyperparameter optimization. The paper uses synthetic or benchmark functions where traditional train/test/validation splits for a fixed dataset are not explicitly provided. |
| Hardware Specification | Yes | All simulations were performed on the same HPC cluster with Intel Xeon 8468 Sapphire at 2.1 GHz. |
| Software Dependencies | Yes | Our evaluation first considers unconstrained and then constrained optimization problems using Bo Torch [5]. To solve B-SUB at each iteration, we use CVXOPT [3] with standard parameters for maximum iterations and tolerance. We follow the Bo Torch tutorial6 which performs inference using Hamiltonian Monte Carlo (HMC) with the NUTS sampler from Pyro [7] and uses log EI [1] as acquisition function. 6Available under the MIT license at https://botorch.org/docs/tutorials/saasbo/ (Bo Torch version v0.15.1). |
| Experiment Setup | Yes | For Baye SQP, we set the hyperparameters δf, δc = 0.2 (unless stated otherwise) and K = d + 1, following Wu et al. [65, Corollary 1]. We furthermore as corse bounds on the lengthscales as ℓi [0.001, 2d] for all baselines and set the noise to a small values i.e., , σ2 = 10 4. The initial length of the trust region is Linit = 0.8, the minimum and maximum length of the trust region are Lmin = 0.57 and Lmax = 1.6, respectively, the number of consecutive failures before the trust region is shrunk is τfail = max{4, d} , the number of consecutive successes before the trust region is expanded is τsucc = 3. |