Pith. sign in

REVIEW 3 major objections 6 minor 42 references

A rebuilt HEBO-style Bayesian optimizer with fixed defaults improves average performance and run-to-run stability on scientific sequential tasks while staying competitive on hyperparameter benchmarks.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-14 10:05 UTC pith:PNPHPBMR

load-bearing objection Solid engineering modernization of HEBO with broad sequential benchmarks and code; the performance claim holds, but the hetero/non-stationarity causal story is unisolated. the 3 major comments →

arxiv 2607.10669 v1 pith:PNPHPBMR submitted 2026-07-12 cs.LG

Modernizing HEBO: a robust Bayesian optimization baseline for practical heteroskedastic and non-stationary problems

L. A. Zhukov , E. V. Shaburova , D. V. Antonets This is my paper
classification cs.LG
keywords Bayesian optimizationheteroskedasticitynon-stationarityoutput warpingGaussian processacquisition functionssequential experimentationmaterials optimization
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.

Bayesian optimization is used for expensive black-box experiments in chemistry and materials science, but standard Gaussian-process surrogates assume constant noise and stationary length scales that real objectives often violate. The authors rebuild the HEBO design philosophy inside a modern framework, revise how output warping is trained and gated, evaluate acquisitions in the original objective space, and search a multi-acquisition Pareto front with a more diversity-aware genetic algorithm. Across synthetic functions, chemistry and 3D-printing emulators, fully experimental reaction grids, needle-in-a-haystack materials problems, and the Bayesmark hyperparameter suite, the resulting model (tidyHEBO) matches or beats original HEBO and a conventional LogEI baseline on average best value while producing tighter confidence bands across repeated runs. Because laboratory budgets are small and a method typically gets only one shot, the authors present the more stable general-purpose optimizer as both a practical tool and a stronger baseline for future work.

Core claim

With fixed default hyperparameters, tidyHEBO delivers competitive-to-superior average performance and visibly tighter run-to-run confidence intervals than original HEBO and conventional LogEI on synthetic, Olympus, fully experimental reaction, and needle-in-a-haystack materials tasks, while remaining competitive on the Bayesmark HPO suite that originally motivated HEBO.

What carries the argument

tidyHEBO: joint training of Yeo-Johnson (or identity) output warping with the Matérn-5/2 Gaussian process, acquisition evaluation on the untransformed objective, and NSGA-III multi-objective search over the acquisition triplet with novelty enforced inside the evolutionary loop.

Load-bearing premise

The reported gains are attributed mainly to better handling of heteroskedastic noise and non-stationary landscapes, yet the paper supplies no ablation that isolates warping from the many simultaneous changes in framework, kernel, acquisition, and search, and does not directly measure those violations on the scientific tasks.

What would settle it

An ablation suite that disables only the gated joint output warping (or only the untransformed acquisition evaluation) on the same Olympus and reaction grids and checks whether the mean curves and confidence-interval widths collapse back toward HEBO or LogEI would falsify the claim that those modeling choices drive the robustness gains.

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

3 major / 6 minor

Summary. The paper introduces tidyHEBO, a BoTorch reimplementation and redesign of HEBO for single-objective sequential Bayesian optimization. It retains HEBO’s high-level design (input/output warping, multi-acquisition Pareto search via a genetic algorithm) while changing the kernel (Matérn 5/2), priors, joint training of Yeo–Johnson output warping with the GP, AF evaluation in the original objective space, LogEI, MC acquisition noise, NSGA-III, novelty constraints inside the GA, and full-history Pareto selection (Table 1). With fixed default hyperparameters, tidyHEBO is evaluated on 6D Ackley/Hartmann, selected Olympus emulators, Shields reaction grids, ZoMBI NIAH materials tasks, and the full Bayesmark suite. The central claim is competitive-to-superior average performance and improved robustness versus original HEBO and LogEI on the scientific tasks, while remaining competitive on Bayesmark (leaderboard score 92.64 vs HEBO 93.34).

Significance. If the empirical package holds, tidyHEBO is a useful, reproducible general-purpose BO baseline for chemistry/materials sequential experimentation, where robustness under small budgets matters as much as mean performance. Strengths include multi-run evaluation (typically 30 runs with bootstrap CIs; 20 full Bayesmark repeats), recovery of HEBO’s published Bayesmark score under sequential evaluation, explicit comparison to domain-specialized methods (EDBO, ZoMBI), and public code. The work is incremental engineering rather than a new theoretical principle, but a carefully modernized, open BoTorch baseline that transfers beyond HPO is of practical value to the applied BO community.

major comments (3)
  1. The title, abstract, and Results and discussion attribute gains primarily to better handling of heteroskedasticity and non-stationarity via warping and related choices. Table 1, however, simultaneously changes framework, kernel, priors, joint vs separate warping training, AF space, LogEI, MC noise, NSGA-III, novelty filtering, and Pareto selection. No ablation isolates warping/joint training (or any single change) from the rest. Without that isolation, the causal story that justifies “modernizing HEBO for practical heteroskedastic and non-stationary problems” is under-supported even if the performance claim is intact. At minimum, ablations on a subset of tasks (e.g., joint vs separate warping; warped vs untransformed AF space; NSGA-II vs III) are needed, or the interpretive claims should be substantially softened to “a modernized HEBO-inspired stack.”
  2. The paper never reports a direct diagnostic that the selected Olympus, reaction, or NIAH tasks actually exhibit heteroskedasticity or non-stationarity (e.g., residual variance vs predicted mean; local length-scale variation; or comparison of stationary vs warped GP fit quality on the emulator/experimental data). Figure 1 and Supplementary Note 2 are illustrative only. Without such evidence, the link from design choices to the claimed problem structure remains an assumption rather than a measured property of the benchmarks used to support the central claim.
  3. Olympus tasks were “manually selected … based on the number of data points … and each problem’s ability to distinguish optimizers by performance” (Results and discussion). Manual selection on discriminability risks optimistic bias for the proposed method. The paper should either (i) report the full Olympus suite (or a pre-registered subset) with the same protocol, or (ii) state the selection criteria and selection date more formally and treat the main figures as illustrative rather than as the primary support for a general-purpose scientific baseline.
minor comments (6)
  1. Table 1 is the most useful summary of differences; consider elevating a short “what changed and why” paragraph earlier so readers do not have to wait until Methods.
  2. Figure 1 panels are labeled a–d but the caption and body text are slightly hard to map; ensure consistent panel order and that “corrected versions” are clearly identified as warped models.
  3. NIAH results use median aggregation without CIs “for clarity”; a short note or supplementary plot with variability would help readers judge robustness claims on those tasks.
  4. Clarify whether the gated warping criterion σ[T(y)]/σ[y] ≥ 1/2 is applied after joint training or only as a post-hoc gate, and whether the model is fully retrained when the gate rejects warping (Methods text is slightly ambiguous).
  5. Minor typography: “performan ce”, “B ayesmark”, duplicated “b” labels in Figure 1, and inconsistent hyphenation of non-stationarity / heteroskedasticity.
  6. Bayesmark recovery of HEBO (~93.34 vs published 93.51) is a strong validation of the experimental setup; state batch-size and sequential protocol differences more prominently in the main text.

Circularity Check

0 steps flagged

No significant circularity: empirical BO methods paper whose performance claims rest on external multi-run benchmarks rather than self-referential derivation.

full rationale

tidyHEBO is presented as an engineering reconstruction of HEBO (input/output warping + multi-AF genetic search) inside BoTorch, with listed implementation changes (Matérn 5/2, joint Yeo-Johnson training, LogEI, NSGA-III, untransformed AF space, etc.; Table 1 and Methods). The load-bearing claims are sequential optimization curves and robustness (mean/median best-so-far with bootstrap CIs) on synthetic functions, Olympus emulators, Shields reaction grids, ZoMBI NIAH emulators, and Bayesmark leaderboard scores. These are evaluated against independent external oracles and baselines (LogEI, original HEBO, Optuna, EDBO, ZoMBI), not quantities defined from the same fitted parameters. The gated warping criterion, λ ~ N(1,0.5^{2}) prior, and joint MLL+Jacobian loss (Supplementary Note 1) are modeling choices whose effect is measured empirically; they do not redefine the reported objective values or force the ranking by construction. No self-citation supplies a uniqueness theorem or ansatz that the central results then merely rename. The paper is therefore self-contained against its benchmarks; any interpretive attribution of gains to heteroskedasticity/non-stationarity is a causal claim outside the circularity patterns.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 1 invented entities

The central claim is empirical performance of a redesigned HEBO-style optimizer. It rests on standard GP/BO machinery, a handful of design choices treated as fixed defaults, and the premise that the chosen benchmarks represent practical heteroskedastic/non-stationary scientific objectives. No new physical entities are postulated; free parameters are modeling hyperparameters held fixed rather than fitted per task.

free parameters (5)
  • Yeo-Johnson λ prior N(1, 0.5²)
    Regularizes gated output warping; chosen by authors and used as a fixed prior across all tasks.
  • Output-warping gate threshold σ[T(y)]/σ[y] ≥ 1/2
    Inherited HEBO-style heuristic that decides whether to keep warping; not derived from the target tasks.
  • MC acquisition samples = 4096
    Fixed Monte Carlo budget for acquisition approximation and LogEI baseline; affects noise and ranking stability.
  • Matérn 5/2 kernel + BoTorch default scaled priors
    Surrogate inductive bias held fixed as a general-purpose choice rather than task-tuned.
  • NSGA-III multi-objective search settings for AF Pareto front
    Genetic algorithm configuration for joint optimization of the acquisition triplet; implementation detail that can change selected candidates.
axioms (5)
  • domain assumption A Gaussian process with Matérn 5/2 kernel plus input/output warping is an adequate surrogate class for the targeted scientific black-box objectives under small budgets.
    Methods section adopts GP + warping as the modeling backbone and attributes gains to heteroskedasticity/non-stationarity handling.
  • domain assumption Acquisition functions should be optimized in the original (untransformed) objective space when outputs are warped.
    Methods cites composite-function BO arguments and makes this a core tidyHEBO change versus HEBO.
  • ad hoc to paper Default hyperparameters held fixed across all tasks are a valid test of generality and robustness.
    Explicit experimental policy in Results; supports the generality claim but is a methodological choice, not a theorem.
  • ad hoc to paper Manually selected Olympus emulators and the chosen reaction/NIAH suites are representative enough to support a general-purpose scientific BO baseline claim.
    Olympus tasks were selected by data size and ability to distinguish optimizers; this selection enters the main empirical argument.
  • standard math Standard BO sequential decision model: maximize f over a bounded design space with noisy observations y = f(x) + ε.
    Stated formally in Methods as the problem setting.
invented entities (1)
  • tidyHEBO independent evidence
    purpose: Name for the specific BoTorch HEBO-inspired optimizer configuration proposed as a robust general-purpose baseline.
    Not a new physical object; it is a software/model package. Independent evidence is the public code and benchmark curves, not an external natural phenomenon.

pith-pipeline@v1.1.0-grok45 · 17907 in / 3382 out tokens · 50757 ms · 2026-07-14T10:05:08.572503+00:00 · methodology

0 comments
read the original abstract

Bayesian optimization is increasingly used to guide data-efficient experimentation in chemistry, materials science, and related laboratory settings, but its practical performance depends strongly on how well surrogate-model assumptions match the geometry and noise structure of the underlying objective. We introduce tidyHEBO, a robust Bayesian optimization model inspired by heteroskedastic evolutionary Bayesian optimization (HEBO) for single-objective, sequential optimization. tidyHEBO reconstructs the HEBO design philosophy in BoTorch and revises surrogate training, output-warping selection, acquisition function evaluation, and Pareto-front search. We benchmarked tidyHEBO on synthetic functions, Olympus emulators, fully experimental reaction-optimization datasets, needle-in-a-haystack (NIAH) materials problems, and Bayesmark hyperparameter optimization tasks. On these tasks tidyHEBO achieved competitive to superior performance and improvement in robustness across repeated optimization runs. We therefore propose tidyHEBO as a practical tool for sequential experimentations and a strong general-purpose benchmark for future Bayesian optimization research.

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

42 extracted references · 18 canonical work pages · 2 internal anchors

  1. [1]

    Benchmarking the performance of Bayesian optimization across multiple experimental materials science doma ins

    Liang Q, Gongora AE, Ren Z, et al. Benchmarking the performance of Bayesian optimization across multiple experimental materials science doma ins. Npj Comput Mater 2021;7:188. https://doi.org/10.1038/s41524-021-00656-9

    doi:10.1038/s41524-021-00656-9 2021

  2. [2]

    Bayesian o ptimization with adaptive surrogate models for automated experimental design

    Lei B, Kirk TQ, Bhattacharya A, et al. Bayesian o ptimization with adaptive surrogate models for automated experimental design. Npj Comput Mater 2021;7:194. https://doi.org/10.1038/s41524-021-00662-x

    work page doi:10.1038/s41524-021-00662-x 2021

  3. [3]

    High Performance Computing for Drug Discovery and Biomedicine

    Heifetz A, editor. High Performance Computing for Drug Discovery and Biomedicine. vol. 2716. New York, NY: Springer US; 2024. https://doi.org/10.1007/978 -1- 0716-3449-3

    doi:10.1007/978 2024

  4. [5]

    Bayesian optimization for chemical products and functional materials

    Wang K, Dowling AW. Bayesian optimization for chemical products and functional materials. Current Opinion in Chem ical Engineering 2022;36:100728. https://doi.org/10.1016/j.coche.2021.100728

    work page doi:10.1016/j.coche.2021.100728 2022

  5. [6]

    Bayesian optimization algorithms for accelerator physics

    Roussel R, Edelen AL, Boltz T, et al. Bayesian optimization algorithms for accelerator physics. Phys Rev Accel Beams 2024;27:084801. https://doi.org/10.1103/PhysRevAccelBeams.27.084801

    doi:10.1103/physrevaccelbeams.27.084801 2024

  6. [7]

    Race to the bottom: Bayesian optimisation for chemical problems

    Wu Y, Walsh A, Ganose AM. Race to the bottom: Bayesian optimisation for chemical problems. Digital Discovery 2024;3:1086 –100. https://doi.org/10.1039/D3DD00234A

    doi:10.1039/d3dd00234a 2024

  7. [8]

    Weighted least -squares in calibration: What difference does it make? Analyst 2007;132:536

    Tellinghuisen J. Weighted least -squares in calibration: What difference does it make? Analyst 2007;132:536. https://doi.org/10.1039/b701696d

    work page doi:10.1039/b701696d 2007

  8. [9]

    Comparison of three weighting schemes in weighted regression analysis for use in a chemistry laboratory

    Jain RB. Comparison of three weighting schemes in weighted regression analysis for use in a chemistry laboratory. Clinica Chimica Acta 2010;411:270 –9. https://doi.org/10.1016/j.cca.2009.11.021

    work page doi:10.1016/j.cca.2009.11.021 2010

  9. [10]

    Validation of quantitative NMR

    Malz F, Jancke H. Validation of quantitative NMR. Journal of Pharmaceutical and Biomedical Analysis 2005;38:813–23. https://doi.org/10.1016/j.jpba.2005.01.043

    work page doi:10.1016/j.jpba.2005.01.043 2005

  10. [11]

    Gaussian processes for machine learning

    Rasmussen CE, Williams CKI. Gaussian processes for machine learning. 3. print. Cambridge, MA: MIT Press; 2006. https://doi.org/10.7551/mitpress/3206.001.0001

    doi:10.7551/mitpress/3206.001.0001 2006

  11. [12]

    HEBO: Pushing The Limits of Sample- Efficient Hyper -parameter Optimisation

    Cowen-Rivers AI, Lyu W, Tutunov R, et al. HEBO: Pushing The Limits of Sample- Efficient Hyper -parameter Optimisation. Journal of Artificial Intelligence Research 2022;74:1269–349. https://doi.org/10.1613/jair.1.13643

    work page doi:10.1613/jair.1.13643 2022

  12. [14]

    Warped Gaussian Processes and Derivative-Based Sequential Designs for Functions with Heterogeneous Variations

    Marmin S, Ginsbourger D, Baccou J, et al. Warped Gaussian Processes and Derivative-Based Sequential Designs for Functions with Heterogeneous Variations. SIAM/ASA J Uncertainty Quantification 2018; 6:991–1018. https://doi.org/10.1137/17M1129179

    work page doi:10.1137/17m1129179 2018

  13. [15]

    An additive Gaussian process regression model for interpretable non -parametric analysis of longitudinal data

    Cheng L, Ramchandran S, Vatanen T, et al. An additive Gaussian process regression model for interpretable non -parametric analysis of longitudinal data. Nat Commun 2019;10:1798. https://doi.org/10.1038/s41467-019-09785-8

    work page doi:10.1038/s41467-019-09785-8 2019

  14. [16]

    Continuous flow Buchwald–Hartwig amination of a pharmaceutical intermediate

    Yaseneva P, Hodgson P, Zakrzewski J, et al. Continuous flow Buchwald–Hartwig amination of a pharmaceutical intermediate. Reaction Chemistry & Engineering 2016;1:229–

    2016

  15. [17]

    https://doi.org/10.1039/C5RE00048C

    work page doi:10.1039/c5re00048c

  16. [18]

    Surrogates: Gaussian process modeling, design, and optimization for the applied sciences

    Gramacy RB. Surrogates: Gaussian process modeling, design, and optimization for the applied sciences. Chapman and Hall/CRC; 2020

    2020

  17. [19]

    Toward real-world automated antibody design with combinatorial Bayesian optimization

    Khan A, Cowen-Rivers AI, Grosnit A, et al. Toward real-world automated antibody design with combinatorial Bayesian optimization. Cell Reports Methods 2023;3 :100374. https://doi.org/10.1016/j.crmeth.2022.100374

    work page doi:10.1016/j.crmeth.2022.100374 2023

  18. [25]

    A Literature Survey of Benchmark Functions For Global Optimization Problems

    Jamil M, Yang X -S. A Literature Survey of Benchmark Functions For Global Optimization Problems. IJMMNO 2013;4:150. https://doi.org/10.1504/IJMMNO.2013.055204

    doi:10.1504/ijmmno.2013.055204 2013

  19. [26]

    Optuna: A Next -generation Hyperparameter Optimization Framework

    Akiba T, Sano S, Yanase T, et al. Optuna: A Next -generation Hyperparameter Optimization Framework. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, New York, NY, USA: Associa tion for Computing Machinery; 2019, p. 2623–31. https://doi.org/10.1145/3292500.3330701

    doi:10.1145/3292500.3330701 2019

  20. [27]

    Imprecise Bay esian optimization

    Rodemann J, Augustin T. Imprecise Bay esian optimization. Knowledge -Based Systems 2024;300:112186. https://doi.org/10.1016/j.knosys.2024.112186

    work page doi:10.1016/j.knosys.2024.112186 2024

  21. [28]

    Revisiting Bayesian Optimization in the light of the COCO benchmark

    Riche RL, Picheny V . Revisiting Bayesian Optimization in the light of the COCO benchmark. Struct Multidisc Optim 2021;64:3063 –87. https://doi.org/10.1007/s00158-021- 02977-1

    work page doi:10.1007/s00158-021- 2021

  22. [29]

    Bayesian Optimization for Anything (BOA): An open -source framework for accessible, user -friendly Bayesian optimization

    Scyphers ME, Missik JEC, Kujawa H, et al. Bayesian Optimization for Anything (BOA): An open -source framework for accessible, user -friendly Bayesian optimization. Environmental Modelling & Software 2024;182:106191. https://doi.org/10.1016/j.envsoft.2024.106191

    doi:10.1016/j.envsoft.2024.106191 2024

  23. [30]

    Practical Bayesian Optimization of Machine Learning Algorithms

    Snoek J, Larochelle H, Adams RP. Practical Bayesian Optimization of Machine Learning Algorithms. Advances in Neural Information Processing Systems, vol. 25, Curran Associates, Inc.; 2012. https://doi.org/10.48550/arXiv.1206.2944

    Pith/arXiv arXiv doi:10.48550/arxiv.1206.2944 2012

  24. [31]

    Vanilla Bayesian Optimization Performs Great in High Dimensions

    Hvarfner C, Hellsten EO, Nardi L. Vanilla Bayesian Optimization Performs Great in High Dimensions. Proceedings of the 41st International Conference on Machine Learning, PMLR; 2024, p. 20793–817. https://doi.org/10.48550/arXiv.2402.02229

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.02229 2024

  25. [32]

    huawei -noah/HEBO: v0.3.4 2022

    Grosnit A, Lyu W, Cowen -Rivers A, et al. huawei -noah/HEBO: v0.3.4 2022. https://doi.org/10.5281/zenodo.7344859

    work page doi:10.5281/zenodo.7344859 2022

  26. [33]

    An Analysis of Transformations

    Box GEP, Cox DR. An Analysis of Transformations. Journal of the Royal Statistical Society Series B: Statistical Methodology 1964;26:211 –43. https://doi.org/10.1111/j.2517- 6161.1964.tb00553.x

    doi:10.1111/j.2517- 1964

  27. [34]

    The Box -Cox Transformation Technique: A Review

    Sakia RM. The Box -Cox Transformation Technique: A Review. The Statistician 1992;41:169. https://doi.org/10.2307/2348250

    doi:10.2307/2348250 1992

  28. [35]

    A new family of power transformations to improve normality or symmetry

    Yeo I-K. A new family of power transformations to improve normality or symmetry. Biometrika 2000;87:954–9. https://doi.org/10.1093/biomet/87.4.954

    doi:10.1093/biomet/87.4.954 2000

  29. [36]

    Warped Gaussian Processes

    Snelson E, Ghahramani Z, Rasmussen C. Warped Gaussian Processes. Advances in Neural Information Processing Systems, vol. 16, MIT Press; 2003, p. 337–44

    2003

  30. [37]

    Unexpected Improvements to Expected Improvement for Bayesian Optimization

    Ament S, Daulton S, Eriksson D, et al. Unexpected Improvements to Expected Improvement for Bayesian Optimization. Advances in Neural Information Processing Systems, vol. 36, Curran Associates, Inc.; 2023, p. 20577 –612. https://doi.org/10.48550/arXiv.2310.20708

    Pith/arXiv arXiv doi:10.48550/arxiv.2310.20708 2023

  31. [38]

    Bayesian Optimization of Composite Functions

    Astudillo R, Frazier PI. Bayesian Optimization of Composite Functions. Proceedings of the 36th International Conference on Machine Learning, PMLR; 2019, p. 354–63. https://doi.org/10.48550/arXiv.1906.01537

    Pith/arXiv arXiv doi:10.48550/arxiv.1906.01537 2019

  32. [39]

    Pymoo: Multi -Objective Optimization in Python

    Blank J, Deb K. Pymoo: Multi -Objective Optimization in Python. IEEE Access 2020;8:89497–509. https://doi.org/10.1109/ACCESS.2020.2990567

    doi:10.1109/access.2020.2990567 2020

  33. [40]

    A fast and elitist multiobjective genetic algorithm: NSGA-II

    Deb K, Pratap A, Agarwal S, et al. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Trans Evol Computat 2002;6:182–97. https://doi.org/10.1109/4235.996017

    doi:10.1109/4235.996017 2002

  34. [41]

    optimization step

    Deb K, Jain H. An Evolutionary Many -Objective Optimization Algorithm Using Reference-Point-Based Nondominated Sorting Approach, Part I: Solving Problems With Box Constraints. IEEE Trans Evol Computat 2014;18:577 –601. https://doi.org/10.1109/TEVC.2013.2281535. Supplementary materials Modernizing HEBO: a robust Bayesian optimization baseline for practical...

    doi:10.1109/tevc.2013.2281535 2014

  35. [42]

    Bayesian Optimization using Deep Gaussian Processes

    Hebbal A, Brevault L, Balesdent M, et al. Bayesian Optimization using Deep Gaussian Processes 2019. https://doi.org/10.48550/arXiv.1905.03350

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1905.03350 2019

  36. [43]

    Olympus: a benchmarking framework for noisy optimization and experiment planning

    Häse F, Aldeghi M, Hickman RJ, et al. Olympus: a benchmarking framework for noisy optimization and experiment planning. Mach Learn: Sci Technol 2021;2:035021. https://doi.org/10.1088/2632-2153/abedc8

    work page doi:10.1088/2632-2153/abedc8 2021

  37. [44]

    Olympus, enhanced: benchmarking mixed- parameter and multi-objective optimization in chemistry and materials science 2023

    Hickman R, Parakh P, Cheng A, et al. Olympus, enhanced: benchmarking mixed- parameter and multi-objective optimization in chemistry and materials science 2023. https://doi.org/10.26434/chemrxiv-2023-74w8d

    work page doi:10.26434/chemrxiv-2023-74w8d 2023

  38. [45]

    Bayesian reaction optimization as a tool for chemical synthesis

    Shields BJ, Stevens J, Li J, et al. Bayesian reaction optimization as a tool for chemical synthesis. Nature 2021;590:89–96. https://doi.org/10.1038/s41586-021-03213-y

    doi:10.1038/s41586-021-03213-y 2021

  39. [46]

    BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization

    Balandat M, Karrer B, Jiang D, et al. BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization. Advances in Neural Information Processing Systems, vol. 33, Curran Associates, Inc.; 2020, p. 21524–38. https://doi.org/10.48550/arXiv.1910.06403

    Pith/arXiv arXiv doi:10.48550/arxiv.1910.06403 2020

  40. [47]

    Fast Bayesian optimization of Needle-in-a-Haystack problems using zooming memory-based initialization (ZoMBI)

    Siemenn AE, Ren Z, Li Q, et al. Fast Bayesian optimization of Needle-in-a-Haystack problems using zooming memory-based initialization (ZoMBI). Npj Comput Mater 2023;9:79. https://doi.org/10.1038/s41524-023-01048-x

    work page doi:10.1038/s41524-023-01048-x 2023

  41. [48]

    Bayesian Optimization is Superior to Random Search for Machine Learning Hyperparameter Tuning: Analysis of the Black-Box Optimization Challenge 2020

    Turner R, Eriksson D, McCourt M, et al. Bayesian Optimization is Superior to Random Search for Machine Learning Hyperparameter Tuning: Analysis of the Black-Box Optimization Challenge 2020. Proceedings of the NeurIPS 2020 Competition and Demonstration Track, vol. 133, PMLR; 2021, p. 3–26. https://doi.org/10.48550/arXiv.2104.10201

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2104.10201 2020

  42. [49]

    Scikit-learn: Machine Learning in Python

    Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research 2011;12:2825–30. https://doi.org/https://doi.org/10.48550/arXiv.1201.0490

    Pith/arXiv arXiv doi:10.48550/arxiv.1201.0490 2011