Pith. sign in

REVIEW 2 major objections 2 minor 48 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

Local methods adapted from standard Bayesian optimization improve preferential optimization in high dimensions by using trust regions and derivatives of the preference model.

2026-06-28 15:52 UTC pith:GRMGUX4K

load-bearing objection The paper adapts trust-region and derivative local search to preferential BO via Laplace-approximated GPs and reports regret gains on high-dim benchmarks, but the approximation's reliability in the claimed regimes is the open question. the 2 major comments →

arxiv 2606.02351 v2 pith:GRMGUX4K submitted 2026-06-01 cs.LG stat.ML

Local Preferential Bayesian Optimization

Johanna Menn , Miriam Kober , Paul Brunzema , David Stenger , Sebastian Trimpe This is my paper
classification cs.LG stat.ML
keywords preferential Bayesian optimizationlocal searchtrust regionGaussian processpairwise preferenceshigh-dimensional optimizationpolicy search
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a family of local preferential Bayesian optimization methods to address the inefficiency of global search in high-dimensional preference learning from pairwise feedback. It adapts trust-region optimization and derivative-informed local search to this setting by exploiting first- and second-order derivatives of the Laplace-approximated Gaussian process posterior. Benchmarks on GP sample paths, standard functions, and policy search tasks demonstrate that these local methods perform especially well in complex landscapes with steep optima and can substantially reduce cumulative regret relative to global preference baselines. A sympathetic reader would care because many practical tuning tasks rely on human pairwise judgments rather than explicit objectives, and extending efficient optimization to higher dimensions makes such tasks more tractable.

Core claim

By transferring trust-region and derivative-informed local search to pairwise preference feedback via the Laplace-approximated GP posterior, local PBO achieves better performance in high-dimensional and complex optimization landscapes than global preference-based methods, with particular gains on steep optima and policy-search problems.

What carries the argument

Local PBO methods adapting trust-region and derivative-informed local search to the preferential setting via the Laplace-approximated GP posterior.

Load-bearing premise

Adapting trust-region and derivative-informed local search from standard BO to the preferential setting via the Laplace-approximated GP posterior will produce reliable performance gains on the chosen benchmarks that generalize to real high-dimensional preference tasks.

What would settle it

A set of benchmark runs on high-dimensional GP sample paths or policy-search tasks in which the local PBO variants fail to produce lower cumulative regret than global preference baselines.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes local preferential Bayesian optimization (PBO) methods that adapt trust-region and derivative-informed local search from standard BO to the pairwise preference setting. The derivative-informed variant extracts gradients and Hessians from a Laplace approximation to the non-Gaussian GP posterior induced by probit or logistic preference likelihoods. Benchmarks on GP sample paths, standard test functions, and policy-search tasks are reported to show that the local methods outperform global PBO baselines and substantially reduce cumulative regret, especially in high-dimensional landscapes with steep optima.

Significance. If the performance claims are robust, the work would address a recognized scalability barrier in PBO by importing local-search machinery that has proven useful in high-dimensional BO. The multi-domain benchmark suite (synthetic GPs, analytic functions, policy search) supplies a reasonable test of generality. No machine-checked proofs or parameter-free derivations are present, but the explicit transfer of trust-region and derivative-informed ideas is a clear methodological contribution.

major comments (2)
  1. [Derivative-informed local search (abstract and methods description)] The derivative-informed local search and trust-region adaptation rely on first- and second-order information extracted from the Laplace-approximated posterior. Because the preference likelihood renders the posterior non-Gaussian, the quadratic fit is known to degrade under sparse data or sharp modes—precisely the high-dimensional steep-optima regime highlighted in the central claim. No diagnostic (e.g., comparison to MCMC or predictive checks on the approximated gradients) is supplied to confirm that the extracted derivatives remain reliable enough to ground the reported regret reductions.
  2. [Benchmark results (abstract)] The benchmark claim that local PBO methods “substantially reduce cumulative regret” relative to global baselines rests on the assumption that the Laplace-based local acquisition and trust-region updates function as intended. If the approximation error is material, the observed gains could be artifacts of the particular benchmark instances rather than a general property of the local formulation.
minor comments (2)
  1. The abstract does not specify the exact form of the pairwise likelihood (probit vs. logistic) or the number of preference queries per iteration; these details affect both the Laplace approximation and the interpretation of the regret curves.
  2. Notation for the local acquisition function and trust-region radius update is introduced without an explicit equation reference, making it difficult to verify how the Laplace derivatives enter the update rules.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's detailed feedback, which highlights important considerations regarding the Laplace approximation in our local PBO methods. We respond to each major comment below and indicate planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Derivative-informed local search (abstract and methods description)] The derivative-informed local search and trust-region adaptation rely on first- and second-order information extracted from the Laplace-approximated posterior. Because the preference likelihood renders the posterior non-Gaussian, the quadratic fit is known to degrade under sparse data or sharp modes—precisely the high-dimensional steep-optima regime highlighted in the central claim. No diagnostic (e.g., comparison to MCMC or predictive checks on the approximated gradients) is supplied to confirm that the extracted derivatives remain reliable enough to ground the reported regret reductions.

    Authors: We acknowledge that the Laplace approximation to the non-Gaussian posterior can have limitations, particularly in regimes with sparse data or sharp modes. While this approximation is commonly used in the PBO literature for its computational efficiency, we agree that additional validation would strengthen the claims. In the revised manuscript, we will include a new subsection providing diagnostics, such as comparisons of Laplace-approximated gradients against MCMC samples on selected benchmark instances, to confirm the reliability of the extracted derivatives in the high-dimensional settings considered. revision: yes

  2. Referee: [Benchmark results (abstract)] The benchmark claim that local PBO methods “substantially reduce cumulative regret” relative to global baselines rests on the assumption that the Laplace-based local acquisition and trust-region updates function as intended. If the approximation error is material, the observed gains could be artifacts of the particular benchmark instances rather than a general property of the local formulation.

    Authors: The referee correctly notes that the performance improvements rely on the effectiveness of the Laplace-based methods. Our multi-domain benchmarks, including GP sample paths, analytic functions, and policy search tasks, show consistent reductions in cumulative regret, which we believe supports the general applicability. To address the concern about potential artifacts, we will expand the discussion in the revised version to include analysis of the approximation quality and its impact on the results, and add the diagnostics mentioned above. revision: yes

Circularity Check

0 steps flagged

No circularity: methods adapt standard GP approximations without definitional reduction

full rationale

The paper's core contribution is the adaptation of trust-region and derivative-informed local search from standard BO to the preferential setting via Laplace approximation of the non-Gaussian posterior induced by pairwise preference likelihoods. No equations or claims in the abstract reduce a reported prediction or regret reduction to a fitted parameter by construction, nor do they rely on self-citations for uniqueness or load-bearing premises. Benchmarks on GP paths, standard functions, and policy search are presented as independent empirical evidence rather than tautological outputs. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no identifiable free parameters, axioms, or invented entities; no equations or implementation details are provided.

pith-pipeline@v0.9.1-grok · 5706 in / 939 out tokens · 32047 ms · 2026-06-28T15:52:32.888020+00:00 · methodology

0 comments
read the original abstract

Bayesian optimization (BO) is a popular and effective approach for tuning expensive, noisy experiments, but requires the formulation of an explicit objective function. Preferential BO (PBO) removes this requirement by learning from pairwise human feedback, yet existing methods struggle to efficiently optimize beyond low- and medium-dimensional problems due to their global search approaches. We address this limitation by developing a family of local PBO methods that transfer key ideas from high-dimensional BO to the preferential setting. In particular, we introduce local PBO methods which adapt trust-region and derivative-informed local search to pairwise preference feedback, where the latter exploits first- and second-order derivatives of the Laplace-approximated GP posterior. Our benchmark on GP sample paths, standard optimization benchmark functions, and policy-search tasks shows that local PBO methods are especially effective in high-dimensional and complex landscapes with steep optima. Compared with global preference-based baselines, they can substantially reduce cumulative regret, making them particularly useful for real-world preference-based optimization tasks such as policy search.

Figures

Figures reproduced from arXiv: 2606.02351 by David Stenger, Johanna Menn, Miriam Kober, Paul Brunzema, Sebastian Trimpe.

Figure 1
Figure 1. Figure 1: Overview of the local methods with 15 evaluations per plot. In GIPBO and PrefSQP, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance on GP sample paths. We plot the final performance (mean ± stdv) after 10d iterations [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Performance on synthetic benchmarks. Best (top) and cumulative (bottom) performance (mean ± stdv). Our local PBO methods achieve higher cumulative performance in high-dimensional settings. 0 80 160 240 320 Iteration 0.00 0.25 0.50 0.75 1.00 fbest( ! ) 1e3 (a) Hopper 0 80 160 240 320 Iteration 0.0 0.4 0.8 1.2 1.6 §f( ! ) 1e5 (b) Hopper 0 250 500 750 1000 Iteration 0 1 2 3 4 fbest( ! ) 1e2 (c) Walker2d 0 250… view at source ↗
Figure 4
Figure 4. Figure 4: Performance on policy-search tasks Hopper and Walker2D. world setting with a crude initial policy [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of gradient steps in GIPBO. (a) The GP posterior is initialized with two data points. (b) A [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: GP posterior mean for parameters x = {x0, x1} and corresponding computation of the two query points to perform the fourth gradient step. C.2.4 Gradient-step normalization We use a squared-exponential kernel on the normalized domain Θ = [0, 1]d , with lengthscales constrained to [0.05, 0.5]. This constraint avoids excessively large local steps, since the gradient direction is normalized with respect to the … view at source ↗
Figure 7
Figure 7. Figure 7: Within-model comparison. Performance over iterations for GP samples between [PITH_FULL_IMAGE:figures/full_fig_p023_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Out-of-model comparison. Performance over iterations for GP samples between [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Final performance boxplots for the different optimization functions. The best performing algorithm has [PITH_FULL_IMAGE:figures/full_fig_p024_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Mujoco Benchmark without warmstarting D.7 Ablations For the different algorithm variants, we perform an ablation in 32d. We conduct the experiments in an out-of-model comparison on 32d GP samples with lengthscales 0.1 and 0.5, on the Levy function in 32d as well as the Styblinsky-Tang 32d test function. For PrefSQP, we ablate the different variants with different acquisition options for the gradient: usin… view at source ↗
Figure 11
Figure 11. Figure 11: PrefSQP variant ablation For TuRPBO, we also compare different acquisition strategies. We compare the perturbed Thompson sampling variant, which is our base TuRPBO version, to expected improvement and qEUBO [1]. On the GP samples TuRPBO performs best, while on the standard benchmarking functions, the qEUBO variant performs equivalently well. 0 80 160 240 320 Iterations 0.0 1.5 3.0 4.5 fbest( ! ) GP 32D, l… view at source ↗
Figure 12
Figure 12. Figure 12: Different Acquisition strategies in TuRPBO [PITH_FULL_IMAGE:figures/full_fig_p025_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Comparison of qEUBO batch size 1, 2 and 4 [PITH_FULL_IMAGE:figures/full_fig_p026_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Noise Ablation 26 [PITH_FULL_IMAGE:figures/full_fig_p026_14.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 10 canonical work pages · 2 internal anchors

  1. [1]

    Raul Astudillo, Zhiyuan Lin, Eytan Bakshy, and Peter I. Frazier. qEUBO: A decision-theoretic acquisition function for preferential Bayesian optimization. InInternational Conference on Artificial Intelligence and Statistics (AISTATS), 25-27 April 2023, Palau de Congressos, Valencia, Spain, volume 206 ofProceedings of Machine Learning Research, pages 1093–1...

    2023

  2. [2]

    Jiang, Samuel Daulton, Benjamin Letham, Andrew Gordon Wilson, and Eytan Bakshy

    Maximilian Balandat, Brian Karrer, Daniel R. Jiang, Samuel Daulton, Benjamin Letham, Andrew Gordon Wilson, and Eytan Bakshy. BoTorch: A framework for efficient Monte-Carlo Bayesian optimization. InAdvances in Neural Information Processing Systems (NeurIPS) 33, 2020

    2020

  3. [3]

    Global optimization based on active preference learning with radial basis functions.Machine Learning, 110(2):417–448, 2021

    Alberto Bemporad and Dario Piga. Global optimization based on active preference learning with radial basis functions.Machine Learning, 110(2):417–448, 2021

    2021

  4. [4]

    Preferential Bayesian optimisation with skew Gaussian processes

    Alessio Benavoli, Dario Azzimonti, and Dario Piga. Preferential Bayesian optimisation with skew Gaussian processes. InProceedings of the Genetic and Evolutionary Computation Conference Companion, pages 1842–1850, 2021

    2021

  5. [5]

    Eric Brochu, Vlad M Cora, and Nando De Freitas. A tutorial on Bayesian optimization of expensive cost functions, with application to active user modeling and hierarchical reinforcement learning.arXiv preprint arXiv:1012.2599, 2010

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  6. [6]

    BayeSQP: Bayesian optimization through sequential quadratic programming

    Paul Brunzema and Sebastian Trimpe. BayeSQP: Bayesian optimization through sequential quadratic programming. InAdvances in Neural Information Processing Systems (NeurIPS), 2025

    2025

  7. [7]

    Preference learning with Gaussian processes

    Wei Chu and Zoubin Ghahramani. Preference learning with Gaussian processes. InProceedings of the 22nd International Conference on Machine Learning (ICML), pages 137–144, 2005

    2005

  8. [8]

    Human-in-the-loop controller tuning using preferential Bayesian optimization.IFAC-PapersOnLine, 58(14):13–18, 2024

    Joao PL Coutinho, Ivan Castillo, and Marco S Reis. Human-in-the-loop controller tuning using preferential Bayesian optimization.IFAC-PapersOnLine, 58(14):13–18, 2024

    2024

  9. [9]

    Safe contextual Bayesian optimization integrated in industrial control for self-learning machines.Journal of Intelligent Manufacturing, 35:885–903, 2024

    Stefano De Blasi, Maryam Bahrami, Elmar Engels, and Alexander Gepperth. Safe contextual Bayesian optimization integrated in industrial control for self-learning machines.Journal of Intelligent Manufacturing, 35:885–903, 2024. doi: 10.1007/s10845-023-02087-3

    work page doi:10.1007/s10845-023-02087-3 2024

  10. [10]

    Savaresi, Sebastian Trimpe, and Simone Formentin

    Giacomo Delcaro, Johanna Menn, David Stenger, Sergio M. Savaresi, Sebastian Trimpe, and Simone Formentin. Twin-in-the-loop observer tuning via gradient-information Bayesian optimization with line search.Control Engineering Practice, 172:106891, 2026. ISSN 0967-

    2026

  11. [11]

    doi: 10.1016/j.conengprac.2026.106891

    work page doi:10.1016/j.conengprac.2026.106891 2026

  12. [12]

    Sequential Preference-Based Optimization

    Ian Dewancker, Jakob Bauer, and Michael McCourt. Sequential preference-based optimization. arXiv preprint arXiv:1801.02788, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  13. [13]

    Consecutive preferential Bayesian optimization, November 2025

    Aras Erarslan, Carlos Sevilla Salcedo, Ville Tanskanen, Anni Nisov, Eero Päiväkumpu, Heikki Aisala, Kaisu Honkapää, Arto Klami, and Petrus Mikkola. Consecutive preferential Bayesian optimization, November 2025. arXiv:2511.05163 [cs]

    work page arXiv 2025

  14. [14]

    Scalable constrained Bayesian optimization

    David Eriksson and Matthias Poloczek. Scalable constrained Bayesian optimization. In International Conference on Artificial Intelligence and Statistics (AISTATS), 2021

    2021

  15. [15]

    Scalable global optimization via local Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 32, 2019

    David Eriksson, Michael Pearce, Jacob Gardner, Ryan D Turner, and Matthias Poloczek. Scalable global optimization via local Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 32, 2019

    2019

  16. [16]

    Minimizing UCB: a better local search strategy in local Bayesian optimization

    Zheyi Fan, Wenyu Wang, Szu H Ng, and Qingpei Hu. Minimizing UCB: a better local search strategy in local Bayesian optimization. InAdvances in Neural Information Processing Systems (NeurIPS), 2024. 10

    2024

  17. [17]

    Cambridge University Press, 2023

    Roman Garnett.Bayesian optimization. Cambridge University Press, 2023

    2023

  18. [18]

    Preferential Bayesian optimization

    Javier González, Zhenwen Dai, Andreas Damianou, and Neil D Lawrence. Preferential Bayesian optimization. InInternational Conference on Machine Learning (ICML), pages 1282–1291. PMLR, 2017

    2017

  19. [19]

    Simulation-aided policy tuning for black-box robot learning.IEEE Transactions on Robotics (TRO), 2025

    Shiming He, Alexander von Rohr, Dominik Baumann, Ji Xiang, and Sebastian Trimpe. Simulation-aided policy tuning for black-box robot learning.IEEE Transactions on Robotics (TRO), 2025

    2025

  20. [20]

    Event-triggered safe Bayesian optimization on quadcopters

    Antonia Holzapfel, Paul Brunzema, and Sebastian Trimpe. Event-triggered safe Bayesian optimization on quadcopters. In6th Annual Learning for Dynamics & Control Conference, pages 1033–1045. PMLR, 2024

    2024

  21. [21]

    Schoellig, and Sebastian Trimpe

    Henrik Hose, Paul Brunzema, Alexander von Rohr, Alexander Gräfe, Angela P. Schoellig, and Sebastian Trimpe. Fine-tuning of neural network approximate MPC without retraining via Bayesian optimization.International Conference on Robot Intelligence Technologies and Applications (RiTA), 2025

    2025

  22. [22]

    User preference meets pareto-optimality in multi-objective Bayesian optimization

    Joshua Hang Sai Ip, Ankush Chakrabarty, Ali Mesbah, and Diego Romeres. User preference meets pareto-optimality in multi-objective Bayesian optimization. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 20246–20254, 2025

    2025

  23. [23]

    Adaptive and safe Bayesian optimization in high dimensions via one-dimensional subspaces

    Johannes Kirschner, Mojmir Mutny, Nicole Hiller, Rasmus Ischebeck, and Andreas Krause. Adaptive and safe Bayesian optimization in high dimensions via one-dimensional subspaces. In International Conference on Machine Learning (ICML), pages 3429–3438. PMLR, 2019

    2019

  24. [24]

    High-dimensional slider-based preferential Bayesian opti- mization with mixed local and global acquisition strategies.IEEE Access, 2024

    Satoshi Koide and Ayano Okoso. High-dimensional slider-based preferential Bayesian opti- mization with mixed local and global acquisition strategies.IEEE Access, 2024

    2024

  25. [25]

    Preference exploration for efficient Bayesian optimization with multiple outcomes

    Zhiyuan Jerry Lin, Raul Astudillo, Peter Frazier, and Eytan Bakshy. Preference exploration for efficient Bayesian optimization with multiple outcomes. InInternational Conference on Artificial Intelligence and Statistics (AISTATS), pages 4235–4258. PMLR, 2022

    2022

  26. [26]

    Duncan Luce

    R. Duncan Luce. Semiorders and a theory of utility discrimination.Econometrica, 24(2): 178–191, 1956

    1956

  27. [27]

    Simple random search of static linear policies is competitive for reinforcement learning

    Horia Mania, Aurelia Guy, and Benjamin Recht. Simple random search of static linear policies is competitive for reinforcement learning. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors,Advances in Neural Information Processing Systems (NeurIPS), volume 31. Curran Associates, Inc., 2018

    2018

  28. [28]

    Choosing a point from the surface of a sphere.The Annals of Mathematical Statistics, pages 645–646, 1972

    George Marsaglia. Choosing a point from the surface of a sphere.The Annals of Mathematical Statistics, pages 645–646, 1972

    1972

  29. [29]

    doi:10.1109/CDC56724.2024.10886720 , booktitle =

    Johanna Menn, Pietro Pelizzari, Michael Fleps-Dezasse, and Sebastian Trimpe. Lipschitz safe Bayesian optimization for automotive control. In2024 IEEE 63rd Conference on Decision and Control (CDC), pages 3782–3788, 2024. doi: 10.1109/CDC56724.2024.10886680

    work page doi:10.1109/cdc56724.2024.10886680 2024

  30. [30]

    Preferential Bayesian optimization with crash feedback.IEEE Robotics and Automation Letters, 11(4):4299–4306, 2026

    Johanna Menn, David Stenger, and Sebastian Trimpe. Preferential Bayesian optimization with crash feedback.IEEE Robotics and Automation Letters, 11(4):4299–4306, 2026. doi: 10.1109/LRA.2026.3665446

    work page doi:10.1109/lra.2026.3665446 2026

  31. [31]

    Projective preferential Bayesian optimization

    Petrus Mikkola, Milica Todorovi ´c, Jari Järvi, Patrick Rinke, and Samuel Kaski. Projective preferential Bayesian optimization. InInternational Conference on Machine Learning (ICML), pages 6884–6892. PMLR, 2020

    2020

  32. [32]

    Local policy search with Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 34:20708–20720, 2021

    Sarah Müller, Alexander von Rohr, and Sebastian Trimpe. Local policy search with Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 34:20708–20720, 2021

    2021

  33. [33]

    Data- efficient autotuning with Bayesian optimization: An industrial control study.IEEE Transactions on Control Systems Technology, 28(3):730–740, 2019

    Matthias Neumann-Brosig, Alonso Marco, Dieter Schwarzmann, and Sebastian Trimpe. Data- efficient autotuning with Bayesian optimization: An industrial control study.IEEE Transactions on Control Systems Technology, 28(3):730–740, 2019

    2019

  34. [34]

    Local Bayesian optimization via maximizing probability of descent

    Quan Nguyen, Kaiwen Wu, Jacob Gardner, and Roman Garnett. Local Bayesian optimization via maximizing probability of descent. InAdvances in Neural Information Processing Systems (NeurIPS), volume 35, 2022

    2022

  35. [35]

    Top-k ranking Bayesian optimization

    Quoc Phong Nguyen, Sebastian Tay, Bryan Kian Hsiang Low, and Patrick Jaillet. Top-k ranking Bayesian optimization. InProceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 9135–9143, 2021. 11

    2021

  36. [36]

    Carl Edward Rasmussen and Christopher K. I. Williams.Gaussian processes for machine learning. Adaptive computation and machine learning. MIT Press, 2006. ISBN 026218253X

    2006

  37. [37]

    Towards practical preferential Bayesian optimization with skew Gaussian processes

    Shion Takeno, Masahiro Nomura, and Masayuki Karasuyama. Towards practical preferential Bayesian optimization with skew Gaussian processes. InInternational Conference on Machine Learning (ICML), pages 33516–33533. PMLR, 2023

    2023

  38. [38]

    Wei-Ting Tang and Joel A. Paulson. CAGES: Cost-aware gradient entropy search for efficient local multi-fidelity Bayesian optimization. In2024 IEEE 63rd Conference on Decision and Control (CDC), pages 1547–1552, December 2024. doi: 10.1109/CDC56724.2024.10886516. ISSN: 2576-2370

    work page doi:10.1109/cdc56724.2024.10886516 2024

  39. [39]

    NeST-BO: Fast local Bayesian optimization via Newton-step targeting of gradient and Hessian information

    Wei-Ting Tang, Akshay Kudva, and Joel Paulson. NeST-BO: Fast local Bayesian optimization via Newton-step targeting of gradient and Hessian information. InThe 29th International Conference on Artificial Intelligence and Statistics (AISTATS), 2026

    2026

  40. [40]

    In: 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

    Maegan Tucker, Myra Cheng, Ellen Novoseller, Richard Cheng, Yisong Yue, Joel W. Burdick, and Aaron D. Ames. Human preference-based learning for high-dimensional optimization of exoskeleton walking gaits. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 3423–3430, 2020. doi: 10.1109/IROS45743.2020.9341416

    work page doi:10.1109/iros45743.2020.9341416 2020

  41. [41]

    Polar: Preference optimization and learning algorithms for robotics.arXiv preprint arXiv:2208.04404, 2022

    Maegan Tucker, Kejun Li, Yisong Yue, and Aaron D Ames. Polar: Preference optimization and learning algorithms for robotics.arXiv preprint arXiv:2208.04404, 2022

    work page arXiv 2022

  42. [42]

    Local Bayesian optimization for controller tuning with crash constraints.at-Automatisierungstechnik, 2024

    Alexander von Rohr, David Stenger, Dominik Scheurenberg, and Sebastian Trimpe. Local Bayesian optimization for controller tuning with crash constraints.at-Automatisierungstechnik, 2024

    2024

  43. [43]

    Individual comparisons by ranking methods.Biometrics Bulletin, 1(6):80–83,

    Frank Wilcoxon. Individual comparisons by ranking methods.Biometrics Bulletin, 1(6):80–83,

  44. [44]

    The behavior and convergence of local Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 36:73497–73523, 2023

    Kaiwen Wu, Kyurae Kim, Roman Garnett, and Jacob Gardner. The behavior and convergence of local Bayesian optimization.Advances in Neural Information Processing Systems (NeurIPS), 36:73497–73523, 2023

    2023

  45. [45]

    Principled preferential Bayesian optimization

    Wenjie Xu, Wenbin Wang, Yuning Jiang, Bratislav Svetozarevic, and Colin Jones. Principled preferential Bayesian optimization. InInternational Conference on Machine Learning (ICML), pages 55305–55336. PMLR, 2024

    2024

  46. [46]

    Improving sample efficiency of high dimensional Bayesian optimization with MCMC

    Zeji Yi, Yunyue Wei, Chu Xin Cheng, Kaibo He, and Yanan Sui. Improving sample efficiency of high dimensional Bayesian optimization with MCMC. In6th Annual Learning for Dynamics & Control Conference, pages 813–824. PMLR, 2024

    2024

  47. [47]

    High-dimensional dueling optimization with preference embedding

    Yangwenhui Zhang, Hong Qian, Xiang Shu, and Aimin Zhou. High-dimensional dueling optimization with preference embedding. InProceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 11280–11288, 2023

    2023

  48. [48]

    experiment-specific

    Max Gero Zimmermann, Johanna Menn, Christopher Ullmann, Viktor Glushych, Thomas Schopphoven, Wilhelm Meiners, Sebastian Trimpe, and Carlo Holly. Bayesian optimization for extreme high-speed laser material deposition.Journal of Laser Applications, 37(4), 2025. 12 A Broader Societal Impacts of the Developed Local PBO Algorithms Our paper proposes three loca...

    2025