arxiv: 2604.01328 · v3 · submitted 2026-04-01 · 💻 cs.LG

Recognition: no theorem link

Efficient and Principled Scientific Discovery through Bayesian Optimization: A Tutorial

Zhongwei Yu , Rasul Tutunov , Alexandre Max Maraval , Zikai Xie , Zhenzhi Tan , Jiankang Wang , Bin Cao , Zijing Li

show 8 more authors

Liangliang Xu Qi Yang Jun Jiang Sanzhong Luo Zhenxiao Guo Tongyi Zhang Haitham Bou-Ammar Jun Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-13 22:03 UTC · model grok-4.3

classification 💻 cs.LG

keywords bayesian optimizationscientific discoverysurrogate modelsacquisition functionsgaussian processescatalysismolecule discoverymaterials science

0 comments

The pith

Bayesian optimization automates the scientific cycle of hypothesis, experiment and refinement using probabilistic models to select informative next steps.

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

The paper frames traditional scientific discovery as an inefficient, ad-hoc process that wastes resources through manual trial and error. It presents Bayesian optimization as a formal framework that treats empirical observations as data for surrogate models, typically Gaussian processes, which serve as evolving hypotheses about an unknown objective function. Acquisition functions then decide which experiment to run next by balancing exploitation of promising regions with exploration of uncertain ones. A sympathetic reader would care because the approach promises to replace guesswork with data-driven decisions in domains such as catalysis, materials design and molecule discovery, leading to fewer experiments and faster progress.

Core claim

Bayesian optimization formalizes scientific discovery as a black-box optimization problem. Surrogate models such as Gaussian processes approximate the unknown objective from observations and provide uncertainty estimates, while acquisition functions quantify the expected benefit of evaluating any candidate point, thereby guiding sequential or batched experiment selection without requiring manual intervention.

What carries the argument

The surrogate model (typically a Gaussian process) paired with an acquisition function inside the Bayesian optimization loop: the surrogate represents current beliefs about the objective function, and the acquisition function scores candidate experiments to balance known performance against uncertainty.

If this is right

Case studies show that BO reduces the number of physical experiments needed in catalysis, organic synthesis and materials science.
Extensions for batched experimentation allow parallel testing while still respecting the acquisition function's recommendations.
Handling of heteroscedasticity lets the framework account for measurement noise that varies across the search space.
Human-in-the-loop integration incorporates domain-expert feedback without breaking the automated selection loop.
Contextual optimization variants handle additional input variables common in real experimental setups.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Fully automated laboratories could run continuous discovery loops with minimal human oversight once surrogate and acquisition components are integrated with robotic systems.
The same loop structure might transfer to non-scientific iterative tasks such as hyperparameter tuning or experimental protocol refinement in other fields.
High-dimensional scientific objectives may require specialized kernels or dimensionality-reduction steps before standard Gaussian-process surrogates become reliable.
Widespread adoption could create standardized experiment-selection logs that improve reproducibility and allow meta-analysis across different labs.

Load-bearing premise

Standard surrogate models such as Gaussian processes can sufficiently approximate the high-dimensional, noisy and often non-stationary objective functions that arise in real scientific problems like catalyst design or molecular synthesis.

What would settle it

A controlled laboratory comparison in which the same discovery task (for example, optimizing a catalyst) is solved once with Bayesian optimization guidance and once with conventional expert-driven or grid search methods, then measuring total experiments required to reach a target performance level.

Figures

Figures reproduced from arXiv: 2604.01328 by Alexandre Max Maraval, Bin Cao, Haitham Bou-Ammar, Jiankang Wang, Jun Jiang, Jun Wang, Liangliang Xu, Qi Yang, Rasul Tutunov, Sanzhong Luo, Tongyi Zhang, Zhenxiao Guo, Zhenzhi Tan, Zhongwei Yu, Zijing Li, Zikai Xie.

**Figure 1.** Figure 1: A four-panel comic illustrates the conceptual journey of translating human scientific intuition into a for [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

**Figure 2.** Figure 2: The illustration of core paradigms of scientific discovery with historical representatives. (i) [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: The AntBO workflow for automated antibody design (Figure 1 of Khan et al. [ [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

**Figure 4.** Figure 4: Workflow of the spatially adaptive active learning strategy. Stage I minimises overpotential via active [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 5.** Figure 5: Scientific discovery as sequential model-based optimisation: iteratively updating a model of the unknown [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗

**Figure 6.** Figure 6: Illustration of surrogate modelling in black-box optimisation. The red curve shows the true objective func [PITH_FULL_IMAGE:figures/full_fig_p024_6.png] view at source ↗

**Figure 7.** Figure 7: ReLU (u) & I(u) Maximising each of these functions implies encouraging their internal expressions to be positive. In the case of the Heaviside function, this corresponds to driving the argument above zero to trigger a nonzero response. For the ReLU function, the goal is to make the argument as large and positive as possible to maximise the acquisition value. Common choices of acquisition functions in Baye… view at source ↗

**Figure 8.** Figure 8: Visualization of the Gaussian Process (GP) surrogate model and acquisition functions used in Bayesian [PITH_FULL_IMAGE:figures/full_fig_p038_8.png] view at source ↗

**Figure 9.** Figure 9: Bayesian Optimisation Loop: At round N, a surrogate model (e.g., a Gaussian process) is fitted to the observed data points ({xt, yt} N t=1). The next point xN+1 is selected by optimising an acquisition function derived from the model posterior. The new observation yN+1 is then evaluated and added to the dataset, iteratively refining the model. The solid brown line depicts the mean prediction of the model, … view at source ↗

**Figure 10.** Figure 10: The optimisation curves (the best input found after each iteration) for the problems with a mathematical [PITH_FULL_IMAGE:figures/full_fig_p060_10.png] view at source ↗

**Figure 11.** Figure 11: Optimisation curves for molecular QED maximisation using descriptor-based GP-UCB and random search. [PITH_FULL_IMAGE:figures/full_fig_p062_11.png] view at source ↗

read the original abstract

Traditional scientific discovery relies on an iterative hypothesise-experiment-refine cycle that has driven progress for centuries, but its intuitive, ad-hoc implementation often wastes resources, yields inefficient designs, and misses critical insights. This tutorial presents Bayesian Optimisation (BO), a principled probability-driven framework that formalises and automates this core scientific cycle. BO uses surrogate models (e.g., Gaussian processes) to model empirical observations as evolving hypotheses, and acquisition functions to guide experiment selection, balancing exploitation of known knowledge and exploration of uncharted domains to eliminate guesswork and manual trial-and-error. We first frame scientific discovery as an optimisation problem, then unpack BO's core components, end-to-end workflows, and real-world efficacy via case studies in catalysis, materials science, organic synthesis, and molecule discovery. We also cover critical technical extensions for scientific applications, including batched experimentation, heteroscedasticity, contextual optimisation, and human-in-the-loop integration. Tailored for a broad audience, this tutorial bridges AI advances in BO with practical natural science applications, offering tiered content to empower cross-disciplinary researchers to design more efficient experiments and accelerate principled scientific discovery.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a well-organized tutorial on standard Bayesian optimization for scientific experiments, but it introduces no new methods, derivations, or results.

read the letter

This paper is a tutorial that frames scientific discovery as an optimization task and explains how Bayesian optimization can guide experiments in areas like catalysis and molecule design. It walks through surrogate models such as Gaussian processes, acquisition functions for balancing exploration and exploitation, and then shows end-to-end workflows with summaries of published case studies. Extensions for batching, heteroscedasticity, and human-in-the-loop setups are included as practical add-ons. The writing stays accessible and the structure is logical, which makes the core ideas usable for readers coming from chemistry or materials backgrounds rather than machine learning. The case-study sections pull together existing applications without re-running them, so they serve as pointers to the literature. The citations track back to established BO work, and the explanations of the standard components line up with what is already known. The main limitation is the lack of novelty. Nothing here derives new bounds, proposes a fresh algorithm, or presents independent experiments that test the framework beyond what the cited papers already showed. The discussion of surrogate models in high-dimensional or noisy scientific settings notes the usual difficulties but relies on common fixes without deeper analysis or new evidence. That leaves the practical claims resting on prior results. This is useful for someone who wants a single document to get started applying BO in the lab without chasing multiple references. It is less relevant for readers already fluent in the BO literature. The thinking is clear and the presentation is honest about its scope as a compilation. A serious editor should send it to peer review for a methods or applications venue where tutorials add value.

Referee Report

1 major / 2 minor

Summary. This tutorial frames scientific discovery as an iterative optimization problem and presents Bayesian Optimization (BO) as a probability-driven framework to automate it. It describes how surrogate models (primarily Gaussian processes) represent evolving hypotheses from observations and how acquisition functions guide experiment selection by balancing exploitation and exploration. The manuscript covers core BO components, end-to-end workflows, extensions for scientific use (batched experiments, heteroscedasticity, contextual optimization, human-in-the-loop), and summarizes case studies in catalysis, materials science, organic synthesis, and molecule discovery.

Significance. If the tutorial's descriptions remain accurate to established BO practice, it offers a useful bridge between machine-learning methods and experimental sciences. The coverage of practical extensions and domain-specific case studies could help non-expert researchers adopt more efficient, less ad-hoc experimental designs, potentially reducing wasted resources in high-cost domains such as catalysis and molecule discovery.

major comments (1)

[Core components] Core components section: the claim that standard surrogate models (Gaussian processes) sufficiently approximate high-dimensional, noisy, and often non-stationary objectives in real scientific domains is load-bearing for the efficacy asserted in the case studies; the text would be strengthened by adding explicit discussion of known failure modes and mitigation strategies rather than relying solely on the general justification.

minor comments (2)

[Abstract] The abstract refers to 'tiered content' for a broad audience, but the manuscript structure does not clearly mark which sections are introductory versus advanced; adding explicit tier indicators or a roadmap would improve accessibility.
[Case studies] Case-study summaries: several applications are described at a high level; including at least one concrete hyperparameter choice or acquisition-function variant per study would aid readers attempting to reproduce or adapt the workflows.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and positive recommendation of minor revision. We address the single major comment below and will revise the manuscript accordingly.

read point-by-point responses

Referee: Core components section: the claim that standard surrogate models (Gaussian processes) sufficiently approximate high-dimensional, noisy, and often non-stationary objectives in real scientific domains is load-bearing for the efficacy asserted in the case studies; the text would be strengthened by adding explicit discussion of known failure modes and mitigation strategies rather than relying solely on the general justification.

Authors: We agree that an explicit discussion of limitations would improve balance and context for the case studies. In the revised manuscript we will expand the Core Components section with a concise subsection on known failure modes of standard Gaussian processes, including the curse of dimensionality for kernel methods, sensitivity to non-stationarity, and performance degradation under high noise. We will also outline common mitigation strategies such as sparse and scalable GP approximations, deep kernel learning, additive or compositional kernels, and hybrid surrogates (e.g., Bayesian neural networks). This addition will be placed after the standard GP description and before the acquisition-function material so that readers encounter the caveats early while the tutorial retains its focus on practical workflows. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper is an expository tutorial that summarizes established Bayesian optimization methods (Gaussian processes, acquisition functions) and their application to scientific domains without performing any new derivations, parameter fits, or theoretical claims. All core components are presented as standard, externally validated techniques rather than being defined or predicted from within the manuscript itself. No self-citation chains, ansatzes, or uniqueness theorems are invoked as load-bearing support for any result, and the content remains self-contained against external benchmarks with no reduction of outputs to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The tutorial rests on the standard modeling assumptions of Bayesian optimization without introducing new free parameters, axioms, or invented entities beyond those already present in the cited literature.

axioms (1)

domain assumption Surrogate models such as Gaussian processes can adequately represent the unknown objective function from finite observations.
Invoked when the tutorial states that surrogate models model empirical observations as evolving hypotheses.

pith-pipeline@v0.9.0 · 5554 in / 1178 out tokens · 44126 ms · 2026-05-13T22:03:58.481871+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

118 extracted references · 118 canonical work pages · 1 internal anchor

[1]

Looping in the Human: Collaborative and Explainable Bayesian Optimization

Masaki Adachi et al. “Looping in the Human: Collaborative and Explainable Bayesian Optimization”. In: Proceedings of The 27th International Conference on Artiﬁcial Intelligence and Statistics . PMLR, Apr. 2024, pp. 505–513

work page 2024
[2]

Using Conﬁdence Bounds for Exploitation-Exploration Trade-offs

Peter Auer. “Using Conﬁdence Bounds for Exploitation-Exploration Trade-offs”. In: Journal of Machine Learning Research 3.Nov (2002), pp. 397–422. ISSN : ISSN 1533-7928

work page 2002
[3]

BoTorch: A framework for efﬁcient Monte-Carlo Bayesian optimization

Maximilian Balandat et al. “BoTorch: A framework for efﬁcient Monte-Carlo Bayesian optimization”. In: Advances in neural information processing systems 33 (2020), pp. 21524–21538

work page 2020
[4]

Max-V alue Entropy Search for Multi-Objective Bayesian Optimization

Syrine Belakaria, Aryan Deshwal, and Janardhan Rao Doppa. “Max-V alue Entropy Search for Multi-Objective Bayesian Optimization”. In: Advances in Neural Information Processing Systems . V ol. 32. Curran Associates, Inc., 2019

work page 2019
[5]

Hyperopt: A Python Library for Optimizing the Hyperparam- eters of Machine Learning Algorithms

James Bergstra, Dan Y amins, and David Cox. “Hyperopt: A Python Library for Optimizing the Hyperparam- eters of Machine Learning Algorithms”. In: Python in Science Conference . Austin, Texas, 2013, pp. 13–19. DOI : 10.25080/Majora-8b375195-003

work page doi:10.25080/majora-8b375195-003 2013
[6]

Algorithms for Hyper-Parameter Optimization

James Bergstra et al. “Algorithms for Hyper-Parameter Optimization”. In: Advances in Neural Information Processing Systems. V ol. 24. Curran Associates, Inc., 2011

work page 2011
[7]

R., Paolini, G

G. Richard Bickerton et al. “Quantifying the Chemical Beauty of Drugs”. In:Nature Chemistry 4.2 (Feb. 2012), pp. 90–98. ISSN : 1755-4330, 1755-4349. DOI : 10.1038/nchem.1243

work page doi:10.1038/nchem.1243 2012
[8]

1977.tb01600.x

G. E. P . Box and D. R. Cox. “An Analysis of Transformations”. In: Journal of the Royal Statistical Society: Series B (Methodological) 26.2 (Dec. 2018), pp. 211–243. ISSN : 0035-9246. DOI : 10.1111/j.2517-6161. 1964.tb00553.x. eprint: https://academic.oup.com/jrsssb/article-pdf/26/2/211/49099371/ jrsssb_26_2_211.pdf. URL : https://doi.org/10.1111/j.2517-6...

work page doi:10.1111/j.2517-6161 2018
[9]

A Mobile Robotic Chemist

Benjamin Burger et al. “A Mobile Robotic Chemist”. In: Nature 583.7815 (July 2020), pp. 237–241. ISSN : 0028-0836, 1476-4687. DOI : 10.1038/s41586-020-2442-2

work page doi:10.1038/s41586-020-2442-2 2020
[10]

On Lower Bounds for Standard and Robust Gaussian Process Bandit Optimiza- tion

Xu Cai and Jonathan Scarlett. On Lower Bounds for Standard and Robust Gaussian Process Bandit Optimiza- tion. May 2021. DOI : 10.48550/arXiv.2008.08757. arXiv: 2008.08757 [stat]

work page doi:10.48550/arxiv.2008.08757 2021
[11]

Active learning accelerates the discovery of high strength and high ductility lead-free solder alloys

Bin Cao et al. “Active learning accelerates the discovery of high strength and high ductility lead-free solder alloys”. In: Materials & Design 241 (2024), p. 112921

work page 2024
[12]

Bgolearn: a Uniﬁed Bayesian Optimization Framework for Accelerating Materials Discovery

Bin Cao et al. “Bgolearn: a Uniﬁed Bayesian Optimization Framework for Accelerating Materials Discovery”. In: arXiv preprint arXiv:2601.06820 (2026)

work page arXiv 2026
[13]

Spatial-adaptive active learning identiﬁes ultra-durable and highly active catalysts for acidic oxygen evolution reaction

Bin Cao et al. “Spatial-adaptive active learning identiﬁes ultra-durable and highly active catalysts for acidic oxygen evolution reaction”. In: Science Bulletin (2025)

work page 2025
[14]

LLM-Enhanced Bayesian Optimization for Efﬁcient Analog Layout Constraint Generation

Guojin Chen et al. LLM-Enhanced Bayesian Optimization for Efﬁcient Analog Layout Constraint Generation . Dec. 2024. DOI : 10.48550/arXiv.2406.05250. arXiv: 2406.05250 [cs]

work page doi:10.48550/arxiv.2406.05250 2024
[15]

Fast Computation of the Multi-Points Expected Improvement with Applications in Batch Selection

Clément Chevalier and David Ginsbourger. “Fast Computation of the Multi-Points Expected Improvement with Applications in Batch Selection”. In: Learning and Intelligent Optimization . Ed. by Giuseppe Nicosia and Panos Pardalos. Berlin, Heidelberg: Springer, 2013, pp. 59–69. ISBN : 978-3-642-44973-4. DOI : 10.1007/ 978-3-642-44973-4_7

work page 2013
[16]

Language-Based Bayesian Optimization Research Assistant (BORA)

Abdoulatif Cissé et al. Language-Based Bayesian Optimization Research Assistant (BORA) . Jan. 2025. DOI : 10.48550/arXiv.2501.16224. arXiv: 2501.16224 [cs]

work page doi:10.48550/arxiv.2501.16224 2025
[17]

On Serendipity in Science: Discovery at the Intersection of Chance and Wisdom

Samantha Copeland. “On Serendipity in Science: Discovery at the Intersection of Chance and Wisdom”. In: Synthese 196.6 (June 1, 2019), pp. 2385–2406. ISSN : 1573-0964. DOI : 10.1007/s11229-017-1544-3

work page doi:10.1007/s11229-017-1544-3 2019
[18]

Cowen-Rivers et al

Alexander I. Cowen-Rivers et al. HEBO: Pushing The Limits of Sample-Efﬁcient Hyperparameter Optimisa- tion. May 2022. DOI : 10.48550/arXiv.2012.03826. arXiv: 2012.03826 [cs]

work page doi:10.48550/arxiv.2012.03826 2022
[19]

Bayesian Optimization Meets Bayesian Optimal Stopping

Zhongxiang Dai et al. “Bayesian Optimization Meets Bayesian Optimal Stopping”. In: Proceedings of the 36th International Conference on Machine Learning . PMLR, May 2019, pp. 1496–1506

work page 2019
[20]

BOA T: Building Auto-Tuners with Structured Bayesian Optimization

V alentin Dalibard, Michael Schaarschmidt, and Eiko Y oneki. “BOA T: Building Auto-Tuners with Structured Bayesian Optimization”. In: Proceedings of the 26th International Conference on World Wide Web . Perth Australia: International World Wide Web Conferences Steering Committee, Apr. 2017, pp. 479–488. ISBN : 978-1-4503-4913-0. DOI : 10.1145/3038912.3052662

work page doi:10.1145/3038912.3052662 2017
[21]

Robust Multi-Objective Bayesian Optimization Under Input Noise

Samuel Daulton et al. Robust Multi-Objective Bayesian Optimization Under Input Noise . June 2022. DOI : 10.48550/arXiv.2202.07549. arXiv: 2202.07549 [cs]

work page doi:10.48550/arxiv.2202.07549 2022
[22]

High-Dimensional Gaussian Process Bandits

Josip Djolonga, Andreas Krause, and V olkan Cevher. “High-Dimensional Gaussian Process Bandits”. In: Ad- vances in Neural Information Processing Systems . V ol. 26. Curran Associates, Inc., 2013. 66 BO Tutorial

work page 2013
[23]

Structure Discovery in Nonparametric Regression through Compositional Kernel Search

David Duvenaud et al. Structure Discovery in Nonparametric Regression through Compositional Kernel Search. May 2013. DOI : 10.48550/arXiv.1302.4922. arXiv: 1302.4922 [stat]

work page doi:10.48550/arxiv.1302.4922 2013
[24]

Scalable Global Optimization via Local Bayesian Optimization

David Eriksson et al. “Scalable Global Optimization via Local Bayesian Optimization”. In:Advances in Neural Information Processing Systems. V ol. 32. Curran Associates, Inc., 2019

work page 2019
[25]

Peter I. Frazier. A Tutorial on Bayesian Optimization. July 2018. DOI : 10.48550/arXiv.1807.02811. arXiv: 1807.02811 [stat]

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.02811 2018
[26]

Sample Efﬁciency Matters: A Benchmark for Practical Molecular Optimization

Wenhao Gao et al. Sample Efﬁciency Matters: A Benchmark for Practical Molecular Optimization . Oct. 2022. DOI : 10.48550/arXiv.2206.12411. arXiv: 2206.12411 [cs]

work page doi:10.48550/arxiv.2206.12411 2022
[27]

Bayesian Optimization with Inequality Constraints

Jacob Gardner et al. “Bayesian Optimization with Inequality Constraints”. In: Proceedings of the 31st Interna- tional Conference on Machine Learning . PMLR, June 2014, pp. 937–945

work page 2014
[28]

GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Accelera- tion

Jacob Gardner et al. “GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Accelera- tion”. In: Advances in Neural Information Processing Systems . V ol. 31. Curran Associates, Inc., 2018

work page 2018
[29]

Bayesian Optimization

Roman Garnett. Bayesian Optimization. Cambridge University Press, 2023

work page 2023
[30]

Kriging Is Well-Suited to Parallelize Optimiza- tion

David Ginsbourger, Rodolphe Le Riche, and Laurent Carraro. “Kriging Is Well-Suited to Parallelize Optimiza- tion”. In: Computational Intelligence in Expensive Optimization Problems . Ed. by Y oel Tenne and Chi-Keong Goh. Berlin, Heidelberg: Springer, 2010, pp. 131–162. ISBN : 978-3-642-10701-6. DOI : 10 . 1007 / 978 - 3 - 642-10701-6_6

work page 2010
[31]

A Multi-points Criterion for Deterministic Par- allel Global Optimization Based on Gaussian Processes

David Ginsbourger, Rodolphe Le Riche, and Laurent Carraro. A Multi-points Criterion for Deterministic Par- allel Global Optimization Based on Gaussian Processes . Report. Mar. 2008

work page 2008
[32]

Decision Making in Medicinal Chemistry: The Power of Our Intuition

Laurent Gomez. “Decision Making in Medicinal Chemistry: The Power of Our Intuition”. In: ACS Medicinal Chemistry Letters 9.10 (Oct. 11, 2018), pp. 956–958. DOI : 10.1021/acsmedchemlett.8b00359

work page doi:10.1021/acsmedchemlett.8b00359 2018
[33]

GLASSES: Relieving The Myopia Of Bayesian Opti- misation

Javier Gonzalez, Michael Osborne, and Neil Lawrence. “GLASSES: Relieving The Myopia Of Bayesian Opti- misation”. In: Proceedings of the 19th International Conference on Artiﬁcial Intelligence and Statistics. PMLR, May 2016, pp. 790–799

work page 2016
[34]

Batch Bayesian Optimization via Local Penalization

Javier Gonzalez et al. “Batch Bayesian Optimization via Local Penalization”. In: Proceedings of the 19th International Conference on Artiﬁcial Intelligence and Statistics . PMLR, May 2016, pp. 648–657

work page 2016
[35]

A Survey and Benchmark of High-Dimensional Bayesian Optimization of Dis- crete Sequences

Miguel González-Duque et al. A Survey and Benchmark of High-Dimensional Bayesian Optimization of Dis- crete Sequences. Nov. 2024. DOI : 10.48550/arXiv.2406.04739. arXiv: 2406.04739 [cs]

work page doi:10.48550/arxiv.2406.04739 2024
[36]

2013 , PAGES =

Robert B. Gramacy and Herbert K. H. Lee. “Optimization Under Unknown Constraints”. In: Bayesian Statis- tics 9 . Ed. by José M. Bernardo et al. Oxford University Press, Oct. 2011, pp. 229–256. ISBN : 978-0-19- 969458-7. DOI : 10.1093/acprof:oso/9780199694587.003.0008

work page doi:10.1093/acprof:oso/9780199694587.003.0008 2011
[37]

Gryfﬁn: An Algorithm for Bayesian Optimization of Categorical V ariables Informed by Expert Knowledge

Florian Häse et al. “Gryfﬁn: An Algorithm for Bayesian Optimization of Categorical V ariables Informed by Expert Knowledge”. In: Applied Physics Reviews 8.3 (Sept. 2021), p. 031406. ISSN : 1931-9401. DOI : 10 . 1063/5.0048164. arXiv: 2003.12127 [stat]

work page arXiv 2021
[38]

Phoenics: A Bayesian Optimizer for Chemistry

Florian Häse et al. “Phoenics: A Bayesian Optimizer for Chemistry”. In: ACS Central Science 4.9 (Sept. 2018), pp. 1134–1145. ISSN : 2374-7943, 2374-7951. DOI : 10.1021/acscentsci.8b00307

work page doi:10.1021/acscentsci.8b00307 2018
[39]

MCMC for V ariationally Sparse Gaussian Processes

James Hensman et al. “MCMC for V ariationally Sparse Gaussian Processes”. In:Advances in Neural Informa- tion Processing Systems. V ol. 28. Curran Associates, Inc., 2015

work page 2015
[40]

Scientiﬁc Method

Brian Hepburn and Hanne Andersen. “Scientiﬁc Method”. In: The Stanford Encyclopedia of Philosophy . Ed. by Edward N. Zalta and Uri Nodelman. Spring 2026. Metaphysics Research Lab, Stanford University, 2026

work page 2026
[41]

Decentralized High-Dimensional Bayesian Optimization With Factor Graphs

Trong Nghia Hoang et al. “Decentralized High-Dimensional Bayesian Optimization With Factor Graphs”. In: Proceedings of the AAAI Conference on Artiﬁcial Intelligence . V ol. 32. Apr. 2018. DOI : 10 . 1609 / aaai . v32i1.11788

work page 2018
[42]

Modular Mechanisms for Bayesian Optimization

Matthew W Hoffman and Bobak Shahriari. “Modular Mechanisms for Bayesian Optimization”. In: Bayesian Optimization in Academia and Industry, NeurIPS 2014 Workshop . 2014

work page 2014
[43]

Cartesian vs. Radial – A Comparative Evaluation of Two Visualization Tools

Frank Hutter, Holger H. Hoos, and Kevin Leyton-Brown. “Sequential Model-Based Optimization for General Algorithm Conﬁguration”. In: Learning and Intelligent Optimization . Ed. by Carlos A. Coello Coello. Berlin, Heidelberg: Springer, 2011, pp. 507–523. ISBN : 978-3-642-25566-3. DOI : 10.1007/978- 3- 642- 25566- 3_40

work page doi:10.1007/978- 2011
[44]

ZINC20—A Free Ultralarge-Scale Chemical Database for Ligand Discovery

John J. Irwin et al. “ZINC20—A Free Ultralarge-Scale Chemical Database for Ligand Discovery”. In: Journal of Chemical Information and Modeling 60.12 (Dec. 2020), pp. 6065–6073. ISSN : 1549-9596, 1549-960X. DOI : 10.1021/acs.jcim.0c00675

work page doi:10.1021/acs.jcim.0c00675 2020
[45]

Jones, Matthias Schonlau, and William J

Donald R. Jones, Matthias Schonlau, and William J. Welch. “Efﬁcient Global Optimization of Expensive Black-Box Functions”. In: Journal of Global Optimization 13.4 (Dec. 1998), pp. 455–492. ISSN : 1573-2916. DOI : 10.1023/A:1008306431147. 67 BO Tutorial

work page doi:10.1023/a:1008306431147 1998
[46]

AntBO: Towards Real-World Automated Antibody Design with Combinatorial Bayesian Opti- misation

Asif Khan et al. AntBO: Towards Real-World Automated Antibody Design with Combinatorial Bayesian Opti- misation. 2022. arXiv: 2201.12570 [q-bio.BM]. URL : https://arxiv.org/abs/2201.12570

work page arXiv 2022
[47]

Contextual Gaussian Process Bandit Optimization

Andreas Krause and Cheng Ong. “Contextual Gaussian Process Bandit Optimization”. In: Advances in Neural Information Processing Systems. V ol. 24. Curran Associates, Inc., 2011

work page 2011
[48]

A Statistical Approach to Some Basic Mine V aluation Problems on the Witwatersrand

D. Krige. “A Statistical Approach to Some Basic Mine V aluation Problems on the Witwatersrand”. In:Journal of The South African Institute of Mining and Metallurgy (Dec. 1951)

work page 1951
[49]

How Useful Is Intermittent, Asynchronous Expert Feedback for Bayesian Optimiza- tion? June 2024

Agustinus Kristiadi et al. How Useful Is Intermittent, Asynchronous Expert Feedback for Bayesian Optimiza- tion? June 2024. DOI : 10.48550/arXiv.2406.06459. arXiv: 2406.06459 [cs]

work page doi:10.48550/arxiv.2406.06459 2024
[50]

Historical Structure of Scientiﬁc Discovery: To the Historian Discovery Is Seldom a Unit Event Attributable to Some Particular Man, Time, and Place

Thomas S. Kuhn. “Historical Structure of Scientiﬁc Discovery: To the Historian Discovery Is Seldom a Unit Event Attributable to Some Particular Man, Time, and Place.” In: Science 136.3518 (June 1962), pp. 760–764. ISSN : 0036-8075, 1095-9203. DOI : 10.1126/science.136.3518.760

work page doi:10.1126/science.136.3518.760 1962
[51]

Thomas S. Kuhn. The Structure of Scientiﬁc Revolutions. 2. ed., enlarged, 21. print. International Encyclopedia of Uniﬁed Science 2,2. Chicago: Univ. of Chicago Press, 1994. ISBN : 978-0-226-45803-8

work page 1994
[52]

A generalized probability density function for double-bounded random pro- cesses

Ponnambalam Kumaraswamy. “A generalized probability density function for double-bounded random pro- cesses”. In: Journal of hydrology 46.1-2 (1980), pp. 79–88

work page 1980
[53]

A New Method of Locating the Maximum Point of an Arbitrary Multipeak Curve in the Presence of Noise

H. J. Kushner. “A New Method of Locating the Maximum Point of an Arbitrary Multipeak Curve in the Presence of Noise”. In: Journal of Basic Engineering 86.1 (Mar. 1964), pp. 97–106. ISSN : 0021-9223. DOI : 10.1115/1.3653121

work page doi:10.1115/1.3653121 1964
[54]

Bandit Algorithms

Tor Lattimore and Csaba Szepesvári. Bandit Algorithms. 1st ed. Cambridge University Press, July 2020. ISBN : 978-1-108-57140-1. DOI : 10.1017/9781108571401

work page doi:10.1017/9781108571401 2020
[55]

Sparse Spectrum Gaussian Process Regression

Miguel Lázaro-Gredilla et al. “Sparse Spectrum Gaussian Process Regression”. In: Journal of Machine Learn- ing Research 11 (2010), pp. 1865–1881

work page 2010
[56]

Bayesian Optimization with Adaptive Surrogate Models for Automated Experimental De- sign

Bowen Lei et al. “Bayesian Optimization with Adaptive Surrogate Models for Automated Experimental De- sign”. In: npj Computational Materials 7.1 (Dec. 2021), p. 194. ISSN : 2057-3960. DOI : 10.1038/s41524- 021-00662-x

work page doi:10.1038/s41524- 2021
[57]

Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks

Chunyuan Li et al. “Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks”. In: Proceedings of the AAAI Conference on Artiﬁcial Intelligence . V ol. 30. Feb. 2016. DOI : 10 . 1609 / aaai . v30i1.10200

work page 2016
[58]

Current Complexity: A Tool for Assessing the Complexity of Organic Molecules

Jun Li and Martin D. Eastgate. “Current Complexity: A Tool for Assessing the Complexity of Organic Molecules”. In: Organic & Biomolecular Chemistry 13.26 (2015), pp. 7164–7176. ISSN : 1477-0520. DOI : 10.1039/C5OB00709G

work page doi:10.1039/c5ob00709g 2015
[59]

Optimize the quantum yield of G-quartet-based circularly polarized luminescence materials via active learning strategy-BgoFace

Tianliang Li et al. “Optimize the quantum yield of G-quartet-based circularly polarized luminescence materials via active learning strategy-BgoFace”. In: Materials Genome Engineering Advances 3.3 (2025), e70031

work page 2025
[60]

Benchmarking the Performance of Bayesian Optimization across Multiple Experimental Materials Science Domains

Qiaohao Liang et al. “Benchmarking the Performance of Bayesian Optimization across Multiple Experimental Materials Science Domains”. In: npj Computational Materials 7.1 (Nov. 2021), p. 188. ISSN : 2057-3960. DOI : 10.1038/s41524-021-00656-9

work page doi:10.1038/s41524-021-00656-9 2021
[61]

Bayesian Epistemology

Hanti Lin. “Bayesian Epistemology”. In: The Stanford Encyclopedia of Philosophy . Ed. by Edward N. Zalta and Uri Nodelman. Summer 2024. Metaphysics Research Lab, Stanford University, 2024

work page 2024
[62]

Large Language Models to Enhance Bayesian Optimization

Tennison Liu et al. Large Language Models to Enhance Bayesian Optimization . Mar. 2024. DOI : 10.48550/ arXiv.2402.03921. arXiv: 2402.03921 [cs]

work page arXiv 2024
[63]

Text Classiﬁcation Using String Kernels

Huma Lodhi et al. “Text Classiﬁcation Using String Kernels”. In: Advances in Neural Information Processing Systems. V ol. 13. MIT Press, 2000

work page 2000
[64]

Active learning in materials science with emphasis on adaptive sampling using uncer- tainties for targeted design

Turab Lookman et al. “Active learning in materials science with emphasis on adaptive sampling using uncer- tainties for targeted design”. In: npj Computational Materials 5.1 (2019), p. 21

work page 2019
[65]

Self-Driving Laboratory for Accelerated Discovery of Thin-Film Materials

B. P . MacLeod et al. “Self-Driving Laboratory for Accelerated Discovery of Thin-Film Materials”. In:Science Advances 6.20 (May 2020), eaaz8867. DOI : 10.1126/sciadv.aaz8867

work page doi:10.1126/sciadv.aaz8867 2020
[66]

Sequential Large Language Model-Based Hyper-parameter Optimiza- tion

Kanan Mahammadli and Seyda Ertekin. Sequential Large Language Model-Based Hyper-parameter Optimiza- tion. Jan. 2025. DOI : 10.48550/arXiv.2410.20302. arXiv: 2410.20302 [cs]

work page doi:10.48550/arxiv.2410.20302 2025
[67]

On the Bayes Methods for Seeking the Extremal Point

J. Mockus. “On the Bayes Methods for Seeking the Extremal Point”. In: IF AC Proceedings V olumes8.1 (Aug. 1975), pp. 428–431. ISSN : 14746670. DOI : 10.1016/S1474-6670(17)67769-3

work page doi:10.1016/s1474-6670(17)67769-3 1975
[68]

BOSS: Bayesian Optimization over String Spaces

Henry Moss et al. “BOSS: Bayesian Optimization over String Spaces”. In: Advances in Neural Information Processing Systems. V ol. 33. Curran Associates, Inc., 2020, pp. 15476–15486

work page 2020
[69]

PFNs4BO: In-Context Learning for Bayesian Optimization

Samuel Müller et al. “PFNs4BO: In-Context Learning for Bayesian Optimization”. In: Proceedings of the 40th International Conference on Machine Learning . PMLR, July 2023, pp. 25444–25470. 68 BO Tutorial

work page 2023
[70]

Murphy.Machine Learning: A Probabilistic Perspective

Kevin P . Murphy.Machine Learning: A Probabilistic Perspective. 4. print. (ﬁxed many typos). Adaptive Com- putation and Machine Learning Series. Cambridge, Mass.: MIT Press, 2013. ISBN : 978-0-262-01802-9

work page 2013
[71]

Radford M. Neal. Bayesian Learning for Neural Networks . Ed. by P . Bickel et al. V ol. 118. Lecture Notes in Statistics. New Y ork, NY: Springer New Y ork, 1996.ISBN : 978-0-387-94724-2. DOI : 10.1007/978-1-4612- 0745-0

work page doi:10.1007/978-1-4612- 1996
[72]

Photocatalytic Water Splitting

Shunta Nishioka et al. “Photocatalytic Water Splitting”. In: Nature Reviews Methods Primers 3.1 (June 2023), p. 42. ISSN : 2662-8449. DOI : 10.1038/s43586-023-00226-x

work page doi:10.1038/s43586-023-00226-x 2023
[73]

Jorge Nocedal and Stephen J. Wright. Numerical Optimization. Second edition. Springer Series in Operations Research and Financial Engineering. New Y ork, NY: Springer, 2006. ISBN : 978-0-387-30303-1

work page 2006
[74]

Bayesian Optimization: A Python Implementation of Global Optimization with Gaussian Processes

Fernando Nogueira. Bayesian Optimization: A Python Implementation of Global Optimization with Gaussian Processes. https://github.com/bayesian-optimization/BayesianOptimization. 2014

work page 2014
[75]

Scikit-learn: Machine Learning in Python

F. Pedregosa et al. “Scikit-learn: Machine Learning in Python”. In: Journal of Machine Learning Research 12 (2011), pp. 2825–2830

work page 2011
[76]

Bayesian Optimization under Mixed Constraints with a Slack-V ariable Augmented La- grangian

Victor Picheny et al. “Bayesian Optimization under Mixed Constraints with a Slack-V ariable Augmented La- grangian”. In: Advances in Neural Information Processing Systems . V ol. 29. Curran Associates, Inc., 2016

work page 2016
[77]

Data Science and Its Relationship to Big Data and Data-Driven Decision Making

Foster Provost and Tom Fawcett. “Data Science and Its Relationship to Big Data and Data-Driven Decision Making”. In: Big Data 1.1 (Mar. 1, 2013), pp. 51–59. ISSN : 2167-6461. DOI : 10.1089/big.2013.1508

work page doi:10.1089/big.2013.1508 2013
[78]

Modern Bayesian Experimental Design

Tom Rainforth et al. Modern Bayesian Experimental Design . Nov. 2023. DOI : 10 . 48550 / arXiv . 2302 . 14545. arXiv: 2302.14545 [stat]

work page arXiv 2023
[79]

Bayesian Optimization of Catalysis With In-Context Learning

Mayk Caldas Ramos et al. Bayesian Optimization of Catalysis With In-Context Learning . May 2025. DOI : 10.48550/arXiv.2304.05341. arXiv: 2304.05341 [physics]

work page doi:10.48550/arxiv.2304.05341 2025
[80]

https://www.rdkit.org/

RDKit: Open-Source Cheminformatics Software. https://www.rdkit.org/

work page

Showing first 80 references.