Pith. sign in

REVIEW 4 major objections 5 minor 50 references

Bayesian Optimization of Pythia8 Tunes

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A Bayesian-optimization search over six Pythia8 parameters produced a tune that fits LEP I data better than the default tune, lowering the fit objective from 3.208 to 2.845.

desk verdict A useful, honest tuning study whose central BOTORCH claim is plausible but not yet robustly established because the improvement is measured on the same noisy runs used for selection. read the letter →

arxiv 2505.11675 v1 pith:ETIGQI5H submitted 2025-05-16 hep-ph hep-exstat.AP

Ali Al Kadhim , Harrison B Prosper , Stephen Mrenna This is my paper · ORCID
classification hep-phhep-exstat.AP
keywords BayesianoptimizationPythia8LundstringmodelhadronizationtuningpartonshowerLEPIdataGaussianprocesssurrogateexpectedimprovement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that Bayesian optimization, a search strategy for expensive black-box functions, can tune the six most influential parameters of the Pythia8 final-state parton shower and hadronization model more effectively than the built-in default tune. The authors build an objective function that compares simulated LEP I electron-positron event-shape and particle-spectrum histograms with data, then minimize it with a Gaussian-process surrogate and an expected-improvement acquisition rule. Their best tune lowers the objective from 3.208 to 2.845, so the tuned model fits the data better by this measure. Two independent Bayesian-optimization implementations converge to essentially the same parameter point, which supports reproducibility of the search. The value of the result is that better automated tuning of hadronization models could improve simulations of proton-proton collisions at the LHC, where the same model is applied.

What carries the argument

The load-bearing machinery is a Bayesian optimization loop: a Gaussian-process surrogate with an ARD Matern kernel models the objective function $O(x)$ of Eq. (5) from evaluations at a Sobol sequence of parameter points, and an expected-improvement acquisition function selects the next point to simulate. The objective itself is a root-mean-square of per-bin $\chi^2$-like terms comparing simulated and measured cross sections across all histograms. The physical model being tuned is the Lund string fragmentation picture, with its fragmentation function $f_{\rm Lund}(z)\propto(1-z)^a z^{-1}\exp(-b m_T^2/z)$, and the six tuned parameters control the Lund $a$ and $b$ values, strangeness suppression, diquark suppression, the strong coupling $\alpha_S(M_Z)$, and the shower transverse-momentum cutoff.

What would settle it

Recompute the objective at the claimed best parameters (aLund=1.836, bLund=1.564, ProbStoUD=0.208, probQQtoQ=0.079, alphaSvalue=0.139, pTmin=1.549) and at the default tune using at least ten million events per point, including bin-to-bin and histogram correlations, and see whether the default tune still scores worse; if the gap closes or reverses, the central claim is false.

Watch

Extended reading notes

Core claim

The central claim is that a six-parameter Bayesian-optimization tune of the Pythia8 Lund string fragmentation and final-state shower yields a better description of LEP I QCD data than the default tune. With parameters StringZ:aLund=1.836, StringZ:bLund=1.564, StringFlav:ProbStoUD=0.208, StringFlav:probQQtoQ=0.079, TimeShower:alphaSvalue=0.139, and TimeShower:pTmin=1.549, the objective function in Eq. (5) evaluates to 2.845, compared with 3.208 for the default tune. The authors also find that two different Bayesian-optimization toolchains converge to nearly the same point, and they caution that the optimum sits at the edge of the allowed aLund range and pairs a large shower cutoff with large Lund a and b parameters, so the improved fit may reflect compensation between shower and hadronization mechanisms rather than a more physical parameter set.

Load-bearing premise

The whole comparison rests on the assumption that the objective function of Eq. (5), computed with only 250,000 simulated events per tune point, ad hoc floors for low-uncertainty bins, and no treatment of bin or histogram correlations, ranks model-data agreement in the same order as a statistically complete comparison would.

Editorial extensions

If this is right

  • If the tuning result is correct, expensive event-generator tuning can be done with far fewer full simulations than traditional polynomial-surrogate methods require.
  • The better fit to LEP I data implies that the universal hadronization model, when applied at the LHC, may need a parameter set different from the default to describe QCD final states.
  • The convergence of two independent implementations of Bayesian optimization to the same point suggests the optimum is not an artifact of one code's acquisition function.
  • Because the best point sits at the boundary of the aLund range and pairs high pTmin with large Lund a and b parameters, the tune should be validated on observables outside the fit before being used as a default.
  • The authors find that using all histograms in the data set is the most important factor in producing a good universal tune, so restricting to a subset of observables would likely weaken the result.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct significance test comparing the 0.36 improvement in $O(x)$ against the directly measured Monte Carlo noise (standard deviation about 0.014 at the default point) would sharpen the claim; the paper's Gaussian-process noise estimate of 1.6 is not the same quantity.
  • The same pipeline could be applied to multi-parton-interaction parameters or to underlying-event tunes, where the objective function is even more expensive, provided the surrogate handles many dimensions.
  • A testable extension is to fix $p_{T,\min}$ at the default value and re-optimize the remaining five parameters; if the fit degrades sharply, the reported tune is exploiting shower-hadronization compensation rather than a genuinely better model.
  • Repeating the tune with several independently seeded data splits or with bin-to-bin correlations would show whether the improvement survives a more faithful uncertainty model.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents a Bayesian optimization (BayesOpt) approach to tuning six Pythia8 parameters (StringZ:aLund, StringZ:bLund, StringFlav:ProbStoUD, StringFlav:probQQtoQ, TimeShower:alphaSvalue, TimeShower:pTmin) against published ALEPH LEP I event-shape and identified-particle distributions. The objective function O(x) in Eq. (5) is a bin-wise root-mean-square chi-like discrepancy, evaluated by running Pythia8 with 250,000 events per parameter point and comparing with Rivet. Two implementations are used: a GPytorch-based expected-improvement optimizer and the BOTORCH qNoisyExpectedImprovement routine. The GPytorch runs do not beat the Monash default (best O = 3.789 in Table II vs. 3.208 for Monash), while the BOTORCH run in Table IV reaches O = 2.845 at aLund = 1.836, bLund = 1.564, ProbStoUD = 0.208, probQQtoQ = 0.079, alphaSvalue = 0.139, pTmin = 1.549. The paper claims this tune fits the ALEPH data better than Monash. Code, data, and a Docker environment are provided for reproducibility.

Significance. If the claimed improvement is robust, this is a useful demonstration that modern Bayesian optimization can handle expensive event-generator tuning with a fairly rich dataset, and the reproducible workflow (GitHub repository, Docker image, configuration files) is a concrete strength. The paper is also valuable for its honest discussion of caveats, including the comparable MC and data statistical precision and the possibility that extreme shower parameters can be compensated by hadronization parameters. However, the central claim rests on a small margin in a noisy, selection-biased objective with ad hoc construction choices, and the two toolkits do not actually converge to the same parameter point in a quantitatively convincing way; the evidence therefore needs strengthening before the improvement over Monash can be considered established.

major comments (4)
  1. [Section VI, Tables II-IV; Algorithm 1, step 13] The headline comparison (O = 2.845 for the BOTORCH tune vs. O = 3.208 for Monash) is based on single 250,000-event evaluations of the same stochastic objective that was used to select the point. Since Algorithm 1 returns the argmin of the noisy observations, the quoted optimum is subject to selection bias, and the reported Monash repeatability (std = 0.014) does not bound the bias at the argmin. The GPytorch-fitted sigma_noise = 1.6 is not reconciled with this small repeatability. An independent re-evaluation of the final tune with larger statistics and multiple seeds, together with an estimate of the selection bias, is needed to support the central claim.
  2. [Section IV, Eq. (5)] The objective function contains ad hoc protections: bins with standard deviation below 10^-3 pb are assigned sigma = 1, and histograms with total cross-section below 0.1 pb are omitted. The paper does not test whether the Monash-versus-BOTORCH ranking is stable under these choices. Given the small improvement margin (0.36 in O), the authors should vary these thresholds and the normalization N, and show that the ordering is preserved.
  3. [Section IV and VI] Correlations between bins and histograms are ignored, although many of the ALEPH observables (e.g., multiplicity distributions d19-d23 and the identified-particle spectra) are kinematically correlated. The sum in Eq. (5) therefore overcounts correlated information for both tunes, but the ranking could still be affected. A correlation-aware treatment, or at least a conservative test such as removing whole correlated groups of histograms, is required before the claimed better fit can be interpreted as a genuine model-data agreement.
  4. [Tables II-IV and Section VII] No uncertainties are reported for the tuned parameters or for O(x). Several parameters lie near the boundary of the allowed ranges (aLund = 1.836/2.0, pTmin = 1.549/2.0), and the GPytorch and BOTORCH tunes disagree substantially in bLund (1.873 vs. 1.564) and pTmin (1.207 vs. 1.549). The statement that both implementations yield essentially the same tune is not quantitatively supported without credible intervals or a sensitivity scan around the reported points.
minor comments (5)
  1. [Table IV] The column header 'NBaysOpt' is a typo; it should be 'NBayesOpt'.
  2. [Reference [27]] Reference [27] appears garbled: 'Thomas A Carlin, Bradley P. Louis' should likely be 'Bradley P. Carlin and Thomas A. Louis'.
  3. [Figures 2 and 3; Table V] The figures do not show all histograms listed in Table V (e.g., d19-d23, d35-d36, d39-d40, d44 are absent); the captions should state which of the 40+ histograms are displayed and why.
  4. [Section I, Goal 2] The stated second goal of comparing the BayesOpt result with Professor/Apprentice tunes is not addressed in the results; no Professor-style tune is fitted or compared, so this goal should either be removed or moved to future work.
  5. [Section VI, last paragraph] The sentence reporting 'the mean of O(x) is 3.19 and its standard deviation is 0.014' does not specify the number of repeated 250,000-event runs used to obtain this estimate; this number should be given.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline improvement is the minimized objective: the claim that the BOTORCH tune fits ALEPH data better is a restatement of Eq. (5) at the argmin point, not an independent validation.

  1. fitted input called prediction [Sec. VI, Tables III-IV; Algorithm 1 step 13; Eq. (5)]
    "13: return ˆ x= arg min x D ... Interestingly, the value of the objective function of the tune obtained with BOTORCH is lower than that of the Monash point, that is, we obtain a tune that fits the ALEPH data better than the Monash tune."

    The tune x̂ is defined by Algorithm 1 as the point with the lowest observed value of O(x) in the data set D, and Eq. (5) is the same O(x) used to evaluate both x̂ and Monash in Tables III-IV. Therefore the statement 'value of the objective function of the tune obtained with BOTORCH is lower than that of the Monash point' is the minimization target restated, not an independent test of fit quality. A successful minimizer of O will, by construction, have O below a non-minimizing baseline; the only non-tautological content is that the search actually found such a point. Since no held-out data or higher-statistics re-evaluation of x̂ is reported, the headline 'fits better' reduces to the fitted objective.

full rationale

The paper is a transparent optimization study with reproducible code, and its methodological statements (GP/BOTORCH mechanics, kernel choices, acquisition functions) are not circular; the citations to the authors' own prior work (Pythia8 manual, Apprentice/BROOD/MLHAD) are contextual, not load-bearing for the headline result. The single significant circularity is in the central claim, not the machinery. The tune x̂ is selected in Algorithm 1, step 13, as 'arg min D' of noisy evaluations of the objective O(x) of Eq. (5), and the evidence quoted for 'fits the ALEPH data better than the Monash tune' is the same objective evaluated at that point (2.845 vs 3.208). Reporting that the argmin of O has smaller O than a baseline is a restatement of the optimization target; it is not an independent prediction. There is one non-tautological element: Monash was not fitted to this objective, and the GPytorch runs actually found no point below Monash, so BOTORCH's improvement is not logically guaranteed. However, the paper provides no held-out data, no cross-validation, and no high-statistics re-evaluation of x̂; the repeatability estimate (std=0.014) was made only at the Monash point. The ad hoc bin floor and histogram cuts in Eq. (5) are not tested for stability, which further weakens the external force of the reported margin, though that is a robustness concern rather than circularity. On balance, the headline claim reduces in part to the fitted objective, so the circularity score is 6 rather than 0-2.

Assumptions & free parameters 9 free parameters · 7 assumptions · 0 invented entities

The paper contributes a fit, not a derivation. It introduces six fitted Pythia8 parameters, plus BayesOpt meta-parameters (acquisition xi, GP noise, run budget) that affect the result. No new physical entities are posited. The key axioms concern model adequacy and the chosen objective; the weakest is that the ad hoc objective faithfully captures agreement with ALEPH data.

free parameters (9)
  • StringZ:aLund = 1.836
    Fitted to ALEPH data by minimizing the objective in Eq. (5); often reaches the upper boundary of 2.0 in BOTORCH iterations, suggesting an edge solution.
  • StringZ:bLund = 1.564
    Fitted alongside aLund to ALEPH data; the pair moves to high values relative to Monash.
  • StringFlav:ProbStoUD = 0.208
    Strangeness suppression parameter fitted to the ALEPH identified-particle spectra.
  • StringFlav:probQQtoQ = 0.079
    Diquark suppression parameter fitted to the ALEPH baryon spectra.
  • TimeShower:alphaSvalue = 0.139
    Strong coupling at M_Z in the final-state shower, fitted because the default was determined by a crude earlier tune.
  • TimeShower:pTmin = 1.549
    Shower cutoff fitted to data; the large value relative to Monash 0.5 forces the hadronization model to supply low-pT hadrons.
  • GP likelihood noise sigma_noise = 1.6 (empirical Bayes)
    Optimized as part of the Gaussian process likelihood; the authors caution it should not be read as the objective noise, since a direct simulation estimate at Monash is 0.014.
  • Expected improvement exploration parameter xi = 0.01
    Taken from Ref. [26] and held fixed; affects how much the search explores versus exploits.
  • BayesOpt budget (NBayesOpt, NSobol, Nrestarts) = NBayesOpt 80-200, NSobol 25-300, Nrestarts 25-5000
    Chosen from toy-problem studies; GPytorch results vary from O=3.789 to 7.763 depending on these settings, so the reported tune depends on them.
assumptions (7)
  • domain assumption The Lund string model, as implemented in Pythia8, can describe final-state hadronization with the six chosen parameters.
    Sec. III; if the model family is inadequate, the fitted parameters absorb model error rather than represent physics.
  • domain assumption Hadronization is universal and factorizable from the perturbative parton shower, so e+e- data can tune a model used for hadron collisions.
    Sec. I; this is the physics rationale for using ALEPH LEPI data for LHC tunes.
  • domain assumption The ALEPH histogram values and uncertainties are correct, and bin-to-bin correlations can be neglected.
    Sec. IV; the objective sums squared residuals per bin without covariance terms.
  • ad hoc to paper The objective function in Eq. (5), with the 10^-3 pb floor and the omission of histograms with total cross-section below 0.1 pb, is a meaningful fit metric.
    Sec. IV; these choices keep any single bin from dominating but are not derived from a full statistical model.
  • standard math The objective is Lipschitz continuous, as required for Bayesian optimization.
    Eq. (7) in Sec. V.
  • domain assumption A Gaussian process with ARD Matern kernel (nu=5/2) and xi=0.01 is an adequate surrogate for the objective.
    Sec. V; the choice follows literature and toy studies, with no proof of adequacy on the actual tune landscape.
  • domain assumption Simulating 250,000 events per tune point gives a sufficiently accurate estimate of the objective.
    Sec. IV; effective data counts reach about 950,000, so Monte Carlo statistical error is not negligible.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bayesian Optimization of Pythia8 Tunes." pith.science (2026). https://pith.science/paper/ETIGQI5H

@misc{pith2026250511675,
  author       = {Pith},
  title        = {Pith review of: Bayesian Optimization of Pythia8 Tunes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ETIGQI5H}},
  note         = {Machine review of arXiv:2505.11675}
}
read the original abstract

A new tune (set of model parameters) is found for the six most important parameters of the Pythia8 final state parton shower and hadronization model using Bayesian optimization. The tune fits the LEPI data from ALEPH better than the default tune in Pythia8. To the best of our knowledge, we present the most comprehensive application of Bayesian optimization to the tuning of a parton shower and hadronization model using the LEPI data.

Figures

Figures reproduced from arXiv: 2505.11675 by the authors.

Figure 1
Figure 1. FIG. 1: Impact of randomly sampling aLund and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: The BO EI tune results using [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 44 canonical work pages

  1. [1]

    Investigate the use of Bayesian optimization for tuning Pythia8 parameters by minimizing an ob- jective function. In principle, Bayesian optimiza- tion is ideally suited for such a problem because of the high computational cost of evaluating the ob- jective function, which requires simulating a large number of events at each tune parameter point

    work page

  2. [2]

    Bayesian Optimization of Pythia8 Tunes

    Assess whether Bayesian-optimized tuning con- verges to the same parameter point as ob- tained with methods such as Professor [7] or Apprentice [8]. The remainder of this paper is organized as follows. In Sec. II we briefly review tuning methods and how our approach differs from them. In Sec. III the six Pythia8 parameters, which are the focus of the tuni...

    work page Pith review arXiv 2025

  3. [3]

    observations

    Tuning with Bayesian Optimization The tuning of an event generator involves generating a sufficient number of events at any given tune point, which is computationally costly. Optimization of a costly objec- tive function is the kind of problem for which Bayesian optimization [17] is, in principle, well suited. Bayesian optimization (BayesOpt) has been att...

    work page

  4. [4]

    number of BayesOpt iterations, NBayesOpt

    work page

  5. [5]

    number of Sobol points NSobol

    work page

  6. [6]

    optimization algorithm for the acquisition function, and

    work page

  7. [7]

    In order to cross-check our results, two different toolk- its—GPytorch [30] and BOTORCH [31]—are employed to carry out the Bayesian-optimization studies, in three stages

    the number of restarts in the optimization of the acquisition function, Nrestarts. In order to cross-check our results, two different toolk- its—GPytorch [30] and BOTORCH [31]—are employed to carry out the Bayesian-optimization studies, in three stages. In stage 1 , an ab initio implementation of BayesOpt was developed using GPytorch with its hyper- param...

    work page

  8. [8]

    • main toy all hyperparams.py — full 108-experiment hyperparameter scan

    Repository and Code The complete analysis is hosted at: https://github.com/AliAlkadhim/BayesOptPythia Key scripts and notebooks live under BayesOpt/src/: • main toy.py — single toy-problem Bayesian- optimization run. • main toy all hyperparams.py — full 108-experiment hyperparameter scan. • validate pythia EI.py — Pythia 8 Bayesian- optimization tune. • p...

    work page

Show all 50 references
  1. [9]

    Docker Environment To avoid dependency issues, we provide a Docker image with all required software such as Pythia8 version 8.309, Rivet version 3.1.9, YODA version 1.9.9, HEPMC version 3.02.06, LHAPDF version 6.5.4, GPytorch, BOTORCH, etc. After installing Docker, pull and ru...

  2. [10]

    Toy Problems

    Running the Experiments a. Toy Problems

  3. [11]

    Edit hyperparameters in BayesOpt/src/configs.py

  4. [12]

    Run a single toy run: python BayesOpt/src/main_toy.py

  5. [13]

    To reproduce the full hyperparameter scan: python BayesOpt/src/main_toy_all_hyperparams.py

  6. [14]

    Generate diagnostic plots by executing post processing toy.ipynb. b. Pythia 8 Tune

  7. [15]

    Configure the tune parameters in configs.py

  8. [16]

    Launch the tune: python BayesOpt/src/validate_pythia_EI.py

  9. [17]

    Produce comparison plots and uncertainty en- velopes via post processing pythia.ipynb. 12

  10. [18]

    For a reasonable run of NBaysOpt = 70 optimiza- tion this amounts to 7.3 hours

    Computational Resources A full Pythia8 tune of 6 parameters× 250, 000 events per point) takes around 374 seconds on a machine with the following specifications: 8th Generation Intel Core i7-855OU Processor, 16 GB DDR4-2400 SDRAM (2 x 8GB). For a reasonable run of NBaysOpt = 70...

  11. [19]

    Collins, Davison E

    John C. Collins, Davison E. Soper, and George F. Ster- man. Factorization of Hard Processes in QCD. Adv. Ser. Direct. High Energy Phys. , 5:1–91, 1989

    1989

  12. [20]

    Barate and others

    R. Barate and others. Studies of quantum chromodynam- ics with the ALEPH detector. 294:1–165, 1998. Number: CERN-PPE-96-186 tex.collaboration: ALEPH

    1998

  13. [21]

    Abreu and others

    P. Abreu and others. Tuning and test of fragmentation models based on identified particles and precision event shape data. Zeitschrift f¨ ur Physik, 73:11–60, 1996. Num- ber: CERN-PPE-96-120 tex.collaboration: DELPHI

    1996

  14. [22]

    I. G. Knowles and others. QCD event generators. In CERN workshop on LEP2 physics (followed by 2nd meet- ing, 15-16 jun 1995 and 3rd meeting 2-3 nov 1995) , June

    1995

  15. [23]

    Carl Edward Rasmussen and Christopher K. I. Williams. Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning) . The MIT Press, 2005

    2005

  16. [24]

    Field and Stephen Wolfram

    Richard D. Field and Stephen Wolfram. A QCD Model for e+ e- Annihilation. Nucl. Phys. B , 213:65–84, 1983

    1983

  17. [25]

    Andersson, G

    B. Andersson, G. Gustafson, G. Ingelman, and T. Sj¨ ostrand. Parton fragmentation and string dynamics. Physics Reports, 97(2):31–145, 1983

    1983

  18. [26]

    Systematic event gen- erator tuning for the LHC

    Andy Buckley, Hendrik Hoeth, Heiko Lacker, Holger Schulz, and Jan Eike von Seggern. Systematic event gen- erator tuning for the LHC. The European Physical Jour- nal C: Particles and Fields , 65:331–357, 2010. arXiv: 0907.2973 [hep-ph] Number: IPPP-09-52, DCPT-104-22, LU-TP-09-...

    2010 arXiv

  19. [27]

    Kowalkowski

    Mohan Krishnamoorthy, Holger Schulz, Xiangyang Ju, Wenjing Wang, Sven Leyffer, Zachary Marshall, Stephen Mrenna, Juliane Muller, and James B. Kowalkowski. Apprentice for Event Generator Tuning, March 2021. arXiv:2103.05748

    2021 arXiv

  20. [28]

    High dimensional parameter tuning for event generators

    Johannes Bellm and Leif Gellersen. High dimensional parameter tuning for event generators. Eur. Phys. J. C , 80(1):54, 2020

    2020

  21. [29]

    Wilkinson, Ahmed Youssef, and Jure Zupan

    Christan Bierlich, Philip Ilten, Tony Menzo, Stephen Mrenna, Manuel Szewc, Michael K. Wilkinson, Ahmed Youssef, and Jure Zupan. Reweighting monte carlo predictions and automated fragmentation variations in pythia 8. August 2023. arXiv: 2308.13459 [hep-ph] Num- ber: FERMILAB-PU...

    2023 arXiv

  22. [30]

    Modeling hadronization using machine learning

    Phil Ilten, Tony Menzo, Ahmed Youssef, and Jure Zupan. Modeling hadronization using machine learning. SciPost Physics, 14:027, 2023. Publisher: SciPost

    2023

  23. [31]

    Neural networks for full phase-space reweighting and param- eter tuning

    Anders Andreassen and Benjamin Nachman. Neural networks for full phase-space reweighting and param- eter tuning. Physical Review D: Particles and Fields , 101(9):091901, 2020. arXiv: 1907.08209 [hep-ph]

    2020 arXiv

  24. [32]

    Austin, Mohan Krishnamoorthy, Sven Leyf- fer, Stephen Mrenna, Juliane M¨ uller, and Holger Schulz

    Anthony P. Austin, Mohan Krishnamoorthy, Sven Leyf- fer, Stephen Mrenna, Juliane M¨ uller, and Holger Schulz. Practical algorithms for multivariate rational approxima- tion. Comput. Phys. Commun. , 261:107663, 2021

    2021

  25. [33]

    BROOD: Bilevel and Robust Optimization and Outlier Detection for Efficient Tuning of High-Energy Physics Event Generators

    Wenjing Wang, Mohan Krishnamoorthy, Juliane Muller, Stephen Mrenna, Holger Schulz, Xiangyang Ju, Sven Leyffer, and Zachary Marshall. BROOD: Bilevel and Robust Optimization and Outlier Detection for Efficient Tuning of High-Energy Physics Event Generators. Sci- Post Phys. Core ...

    2022

  26. [34]

    Event generator tunes obtained from underlying event and multiparton scat- tering measurements

    Vardan Khachatryan and others. Event generator tunes obtained from underlying event and multiparton scat- tering measurements. The European Physical Jour- nal C: Particles and Fields , 76(3):155, 2016. arXiv: 1512.00815 [hep-ex] Number: CMS-GEN-14-001, CERN- PH-EP-2015-291 tex...

    2016 arXiv

  27. [35]

    Extraction and validation of a new set of CMS PYTHIA8 tunes from underlying-event measurements

    Albert M Sirunyan et al. Extraction and validation of a new set of CMS PYTHIA8 tunes from underlying-event measurements. Eur. Phys. J. C , 80(1):4, 2020

    2020

  28. [36]

    Adams, and Nando de Freitas

    Bobak Shahriari, Kevin Swersky, Ziyu Wang, Ryan P. Adams, and Nando de Freitas. Taking the human out of the loop: A review of bayesian optimization. Proceedings of the IEEE , 104(1):148–175, 2016

    2016

  29. [37]

    Event generator tuning using Bayesian optimization

    Philip Ilten, Mike Williams, and Yunjie Yang. Event generator tuning using Bayesian optimization. JINST, 12(04):P04028, 2017

    2017

  30. [38]

    Chew and Steven C

    Geoffrey F. Chew and Steven C. Frautschi. Regge trajec- tories and the principle of maximum strength for strong interactions. Phys. Rev. Lett. , 8:41–44, Jan 1962

    1962

  31. [39]

    Hadronic structure on the light front

    Edward Shuryak and Ismail Zahed. Hadronic structure on the light front. ii. qcd strings, wilson lines, and poten- tials. Phys. Rev. D , 107:034024, Feb 2023

    2023

  32. [40]

    Preuss, Torbj¨ orn Sj¨ ostrand, Peter Skands, Marius Utheim, and Rob Verheyen

    Christian Bierlich, Smita Chakraborty, Nishita De- sai, Leif Gellersen, Ilkka Helenius, Philip Ilten, Leif L¨ onnblad, Stephen Mrenna, Stefan Prestel, Chris- tian T. Preuss, Torbj¨ orn Sj¨ ostrand, Peter Skands, Marius Utheim, and Rob Verheyen. A comprehensive guide to the phy...

    2022

  33. [41]

    A. J. Booker, J. E. Dennis, P. D. Frank, D. B. Serafini, V. Torczon, and M. W. Trosset. A rigorous framework for optimization of expensive functions by surrogates. Struc- tural optimization, 17(1):1–13, Feb 1999

    1999

  34. [43]

    Marc G. Genton. Classes of kernels for machine learning: a statistics perspective. J. Mach. Learn. Res. , 2:299–312, March 2002

    2002

  35. [44]

    Jones, Matthias Schonlau, and William J

    Donald R. Jones, Matthias Schonlau, and William J. Welch. Efficient global optimization of expensive black- box functions. Journal of Global Optimization , 13:455– 492, 1998

    1998

  36. [45]

    Practical bayesian optimiza- tion

    Daniel James Lizotte. Practical bayesian optimiza- tion. phd, University of Alberta, CAN, 2008. ISBN: 9780494463659

    2008

  37. [46]

    Thomas A Carlin, Bradley P. Louis. Empirical Bayes: Past, Present, and Future . 2002

    2002

  38. [47]

    On the distribution of points in a cube and the approximate evaluation of integrals

    I.M Sobol’. On the distribution of points in a cube and the approximate evaluation of integrals. USSR Computa- tional Mathematics and Mathematical Physics , 7(4):86– 112, 1967

    1967

  39. [48]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  40. [49]

    GPyTorch: Blackbox matrix-matrix gaussian process inference with GPU acceleration

    Jacob R Gardner, Geoff Pleiss, David Bindel, Kilian Q Weinberger, and Andrew Gordon Wilson. GPyTorch: Blackbox matrix-matrix gaussian process inference with GPU acceleration. In Advances in neural information processing systems, 2018

    2018

  41. [50]

    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, December

  42. [1995]

    arXiv: hep-ph/9601212

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.