Pith. sign in

REVIEW 3 major objections 6 minor 42 references

Fast linear optics models plus genetic algorithms raise CERN North Area beam rates by tens of percent and cut halo five-fold.

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

2026-07-31 09:28 UTC pith:NHCFHO2T

load-bearing objection Real 2025 beam gains on H4/M2 from Xsuite+NSGA-II optics; relative improvements hold, absolute H4 rates still need a hand-tuned offset. the 3 major comments →

arxiv 2607.28370 v1 pith:NHCFHO2T submitted 2026-07-30 physics.acc-ph

Fast optics-based modeling enabling large-scale optimization of the H4 and M2 beamlines in the CERN SPS North Area

Giovanni Dal Maso , Dipanwita Banerjee , Nikolaos Charitonidis This is my paper
classification physics.acc-ph PACS 29.27.Eg29.27.Fh41.85.Ja41.75.Ht
keywords secondary beamlinesoptics optimizationgenetic algorithmsXsuiteBDSIMH4 beamlineM2 beamlinemuon beams
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.

Secondary beamlines at CERN feed many fixed-target experiments and test beams, but their dozens of independently powered magnets create a high-dimensional optics problem that is hard to optimize by hand. This paper shows that a fast, linear-optics tracker is accurate enough to drive multi-objective genetic searches on ordinary workstation hardware, after which only the final candidates need expensive full Monte-Carlo checks. Applied to the H4 electron line serving NA64, the search produced new quadrupole settings that raise the delivered electron rate by 30 percent while shrinking beam halo from 5 percent to 1 percent. On the longer M2 line the same workflow nearly doubles the electron rate and improves muon transmission by two-thirds, both results confirmed in 2025 beam time. The practical message is that large-scale optics redesign of multi-purpose secondary lines is now routine rather than exceptional, giving experiments higher statistics inside the same beam schedule.

Core claim

A two-stage genetic optimization driven by a fast Xsuite optics model, later validated by BDSIM and by 2025 measurements, yields new H4 electron optics that increase the NA64 electron rate by 30 percent and reduce beam-related background five-fold, and new M2 optics that raise muon transmission by 67 percent and nearly double the electron rate.

What carries the argument

The two-stage NSGA-II search (mono-objective rate maximization on the full lattice, then multi-objective rate-plus-phase-space optimization on the downstream section) run on a linear Xsuite model that includes biased hadron decays and sliced absorber interactions; high-fidelity BDSIM is used only to generate intermediate distributions and to benchmark final candidates.

Load-bearing premise

The fast linear tracker must stay faithful enough that the quadrupole and scraper settings it ranks highest remain near-optimal once real material, vacuum imperfections and alignment errors are restored.

What would settle it

A controlled beam test that installs the published optimized quadrupole and scraper currents and measures both absolute particle rate and halo fraction at the experimental stations; if the measured gains fall well below the claimed 30 percent / 67 percent figures after ordinary steering corrections, the claim fails.

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

If this is right

  • NA64 can collect the same electron statistics in roughly 25 percent less beam time, or push sensitivity with the extra rate.
  • MUonE electromagnetic-calorimeter calibrations can be performed with nearly twice the previous electron flux.
  • Muon experiments on M2 gain a two-thirds rate increase without hardware changes, provided radiation limits allow it.
  • The same workstation-scale workflow can be reapplied whenever target, collimator or vacuum conditions change.
  • Vacuum upgrades already planned for M2 are expected to compound the electron-rate gains further.

Where Pith is reading between the lines

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

  • Once the method is routine, other multi-purpose secondary lines (East Area, Fermilab Test Beam, etc.) become natural next targets for the same two-stage search.
  • The need to hand-tune a 200 µrad steering offset to match absolute H4 rates suggests that a modest online alignment or initial-condition layer could close the remaining absolute-rate gap without slowing the optimizer.
  • Radiation-protection ceilings, not optics, will likely set the next practical limit on M2 muon intensity; the optics gains therefore shift the bottleneck rather than remove 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 manuscript presents a two-stage multi-objective optimization (NSGA-II with TPE warm-up) of the CERN SPS North Area H4 and M2 secondary beamlines, driven by fast Xsuite optics tracking and validated against high-fidelity BDSIM Monte Carlo and 2025 beam measurements. For H4 electrons to NA64, an unlike-polarity quadrupole solution increases momentum acceptance and delivers a measured ~30% higher electron rate with beam halo reduced from 5% to 1%. For M2, optimized muon optics plus scraper apertures yield a measured 67% muon-rate increase (optics and scrapers each ~30%), and a separate six-quadrupole electron optimization nearly doubles the electron rate. The central methodological claim is that linear optics models (with limited BeamInteraction extensions for hadron decay and Be absorbers) are fast enough for tens of thousands of evaluations on a standard workstation while remaining faithful enough that Pareto candidates remain near-optimal under full physics.

Significance. If the reported gains hold under routine operation, the work has direct operational value for NA64, MUonE, and other North Area users, and it provides a transferable workflow for high-dimensional secondary-beamline optics. Strengths include: (i) end-to-end experimental deployment and measurement of the optimized settings (Fig. 7, Fig. 8, Tables IV–V); (ii) quantitative Xsuite–BDSIM vacuum benchmarks at the few-percent level on rate and second moments for the very candidates tested (Table I); (iii) explicit speed-up numbers (Table VI: ~10^3 for H4, ~10^2 for M2) that substantiate the “standard hardware” claim; and (iv) a practical, modular constraint formulation that steers the same optimizer toward different experiments. The combination of measured rate/halo gains with a documented fast-model pipeline is a solid contribution to accelerator operations and beamline design.

major comments (3)
  1. [§III.A.1, Fig. 7] §III.A.1 and Fig. 7: absolute H4 rates match measurement only after including the full material budget and a hand-inserted 200 µrad initial-direction offset. Relative gains then agree within the stated ~10% systematic, which supports the strongest claim. However, the paper does not show whether the optimized unlike-polarity lattice is more or less sensitive to steering/alignment than the nominal high-transmission optics. A short differential scan (or BDSIM mis-steering study) of rate and halo versus initial angle for both lattices would establish that the 30% gain and five-fold halo reduction are robust under realistic commissioning errors, not only under the single offset that restores absolute rates.
  2. [§III, §III.C, Table VI] §III and §III.C: the methodological claim that Xsuite enables “large-scale” multi-objective optimization on standard hardware is central, yet the manuscript does not report population size, number of generations/trials, total evaluations, or wall-clock time for the full mono- and multi-objective runs (only per-particle times in Table VI and “8 cores”). Without these numbers, reproducibility of the workflow and an independent assessment of search adequacy in the 18- and 33-dimensional spaces are limited. Adding a brief table or paragraph with Optuna/NSGA-II settings and total evaluations would close this gap without changing the physics conclusions.
  3. [§III.B, Table IV, Table V, Abstract] §III.B (M2 muon multi-objective stage) and Table V: the 67% measured muon-rate gain combines optimized optics with optimized scraper apertures that are explicitly looser and MUonE-specific (Table IV; text notes COMPASS/AMBER need tighter halo). The intermediate configurations already separate ~30% optics vs ~30% scrapers, which is good. The paper should state more clearly in the abstract/conclusions that the full 67% figure is not a drop-in replacement for halo-critical users, and ideally quote a halo or scraper-transmission metric for the optimized apertures so that other experiments can judge applicability.
minor comments (6)
  1. [§III] Section heading “III. OPTIMIZA TION” contains a spurious space (and similar minor spacing artifacts appear elsewhere in the source). Clean for production.
  2. [§II.A.1] Synchrotron-radiation energy loss in H4 is neglected (§II.A.1). A one-sentence estimate of the fractional energy loss at 100 GeV/c over the line, or a statement that strengths are scaled in operation but fixed here, would reassure readers that the omission does not bias the optimized gradients.
  3. [Table I] Table I caption and text: “optimized-like” / “optimized-unlike” naming is clear in context but could be defined once in the table footnote for readers who land on the table first.
  4. [Figs. 3, 5, 6] Fig. 3, 5, 6: R-matrix envelopes are informative; adding the experiment location (s = 544 m for H4, s = 1079 m for M2) as a vertical marker on all optics figures would ease comparison with the FoM evaluation points.
  5. [Abstract, §III.B.2] Abstract says “nearly two-fold increase in electron rate” for M2; body reports 82% (§III.B.2). Align the abstract wording with the measured 82% (or “nearly doubled”) for consistency.
  6. [References, §III] Reference list and tool citations (Xsuite, BDSIM, Optuna, pyg4ometry) are appropriate; consider citing the specific NSGA-II/OptunaHub sampler versions used for full reproducibility.

Circularity Check

0 steps flagged

No circularity: optimization objectives are independent Monte-Carlo figures of merit, validated by external BDSIM benchmarks and 2025 beam measurements.

full rationale

The paper's central claims are empirical performance gains from multi-objective genetic optimization of quadrupole and scraper settings. Rate, beam size, divergence, and skewness are evaluated on independent particle samples in a fast Xsuite tracker; candidate solutions are then re-evaluated in high-fidelity BDSIM and finally measured on the real H4 and M2 beamlines. Nothing is defined in terms of a later-reported prediction, no parameter is fitted to data and then re-presented as a forecast, and self-citations are only to facility descriptions, tool papers (Xsuite, BDSIM, Optuna, NSGA-II), and prior experimental setups. The measured 30 % electron-rate increase and five-fold halo reduction at NA64, the 67 % muon-rate increase on M2, and the ~82 % M2 electron-rate increase are external experimental outcomes, not algebraic consequences of the optimizer's inputs. Absolute-rate discrepancies that require material budget and a steering offset are acknowledged openly and do not close any definitional loop. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central empirical claims rest on standard accelerator tracking assumptions, a small set of hand-chosen optimization constraints and breaking points, and the fidelity of a simplified decay/absorber model. No new physical entities are postulated; free parameters are algorithmic and operational choices rather than fitted constants that define the result.

free parameters (5)
  • H4 multi-objective constraints (spot <20 mm, STD(x)<5.6 mm, STD(y)<15 mm, acceptance >7.8 %) = >7.8 % acceptance, 1.3× nominal sizes
    Numerical thresholds chosen to bias the Pareto search toward ‘strict improvement’ over nominal; the 15 % acceptance floor is explicitly called arbitrary.
  • M2 mono- and multi-objective phase-space and skewness cuts = spot <20–30 mm, div <0.7–1 mrad, g1 bounds
    STD limits and Fisher-Pearson skewness bounds (g1<0.2/0.1) are set by hand to discard asymmetric solutions.
  • Breaking points Q13 (H4) and Q21 (M2) = Q13 / Q21
    Upstream quadrupoles frozen after mono-objective stage; choice reduces dimensionality but is not unique.
  • Hadron-decay bias factor and 20 m BeamInteraction spacing = bias 10 (opt) / 100 (bench), Δs=20 m
    Variance-reduction parameters of the fast muon model; verified a posteriori but still free modeling choices.
  • 200 µrad initial electron direction offset (H4 absolute-rate match) = 200 µrad
    Ad-hoc steering error introduced post-hoc so BDSIM matches commissioning data; not predicted a priori.
axioms (4)
  • domain assumption Linear R-matrix optics plus thin-lens quadrupoles suffice to rank candidate settings for secondary beams when final candidates are re-checked with full Monte Carlo.
    Core modeling premise of the entire Xsuite optimization loop (Sec. III).
  • domain assumption Biased hadron decay at discrete 20 m stations plus sliced continuous-energy-loss/scattering in Be absorbers reproduces muon phase space to a few percent.
    Stated and cross-checked against unbiased BDSIM (Sec. III.B.1).
  • domain assumption NSGA-II with TPE warmup converges to useful Pareto fronts in 18- and 33-dimensional magnet spaces on 8-core workstations.
    Empirical algorithmic premise; no convergence proof offered.
  • ad hoc to paper Synchrotron-radiation energy loss in H4 can be neglected for the present study.
    Explicitly stated in Sec. II.A.1; operational practice scales magnets for it.

reviewed 2026-07-31 · how reviews work

0 comments
read the original abstract

The CERN multi-purpose secondary beamlines are invaluable facilities offering a rich and diverse physics program to the international particle physics community, providing many different particle species and beam characteristics to fixed-target experiments and test-beam users. A number of magnetic elements are required to enable such flexibility, posing a significant challenge in the design and optimization of the beamline optics. In this work, we present the optimization of the H4 and M2 beamlines of the CERN SPS North Area using genetic algorithms. The aim of this study is to demonstrate the effectiveness of fast optics-based tracking models (here Xsuite), to enable large-scale, multi-objective optimization on standard computing hardware. The model is benchmarked and validated with high-fidelity Monte Carlo simulations using BDSIM and cross-validation of the results with experimental measurements. The optimized optics, applied to the H4 electron configuration for NA64, achieved a 30% increase in available electron rate and a five-fold reduction in beam-related background. In the M2 line, the optimized configurations yielded a nearly two-fold increase in electron rate and a 67% improvement in muon transmission.

Figures

Figures reproduced from arXiv: 2607.28370 by Dipanwita Banerjee, Giovanni dal Maso, Nikolaos Charitonidis.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of the CERN accelerator complex. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Schematic layout of the H4 beamline starting from the T2 target station. Dipole magnets are shown in blue, quadrupoles [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. High-transmission optics of the H4 beamline starting from the T2 target center. The R-matrix transport elements [ [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Schematic layout of the M2 beamline from the primary target (T6) towards the EHN2 experimental hall. The hadron [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Nominal optics of the M2 beamline for muon operation along the hadron section (top) and the muon section (bottom). [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Electron rates at the last scintillating detector before the NA64 experiment ( [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Beam spot at the exit of the NA64 vacuum vessel ( [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

42 extracted references · 5 linked inside Pith

  1. [1]

    A schematic overview of the beam- line layout is shown in Fig

    Layout and optical design The H4 beamline is a high-energy, high-resolution sec- ondary beamline served by the T2 target station in the CERN North Area. A schematic overview of the beam- line layout is shown in Fig. 2. The beam composition can be controlled by inserting material at dedicated loca- tions along the beamline. The electron beam is produced fr...

  2. [2]

    Thethree modeswere de- veloped to cover different experimental needs and all three are regularly used today depending on the experiment being served

    Nominal operating modes The H4 beamline is currently operated in three main optical configurations [ 17], selected exclusively through changes in the quadrupole settings: high-transmission, high-resolution, and filter mode. Thethree modeswere de- veloped to cover different experimental needs and all three are regularly used today depending on the experime...

  3. [3]

    Simulation input In this study, we focus on the 100 GeV/c electron beam delivered to the NA64 experiment [16]. The input particle distributions for both the Xsuite andBDSIMmodels of the H4 beamline were generated using aBDSIMsimu- lation of the T2 target area with 10 8 protons on target (PoT), selecting the largest XTAX aperture. To reduce the variance of...

  4. [4]

    The beamline layout is shown in Fig

    Layout and beam characteristics The M2 beamline is a long, high-energy secondary beamline in the CERN North Area, served by the T6 target station and designed to deliver hadron, electron, and muon beams to experiments located in the EHN2 experimental hall. The beamline layout is shown in Fig. 4. The momentum selection, when operating the muon beam, is pro...

  5. [5]

    The de- cay channel is tuned to transport hadrons at a higher momentum, while the downstream section is set to ap- proximately 93 % of the hadron momentum

    Muon beam production For muon operation, the beamline is configured such that the hadron decay section and the downstream muon section operate at different central momenta. The de- cay channel is tuned to transport hadrons at a higher momentum, while the downstream section is set to ap- proximately 93 % of the hadron momentum. This choice represents a com...

  6. [6]

    Electron beam operation In contrast with H4, the XTAX collimator in M2 is located downstream of six quadrupole magnets and the first horizontal bend, which provides the initial momen- tum selection. Therefore, the production of secondary electron beams in M2 cannot rely on the conversion of a collimated photon beam, since there is no direct line of sight ...

  7. [7]

    For the muon beam studies, a hadron beam file was prepared by impinging 10 7 protons on target

    Simulation input The input particle distributions for both the Xsuite andBDSIMmodels of the M2 beamline were generated with aBDSIMsimulation of the T6 target area. For the muon beam studies, a hadron beam file was prepared by impinging 10 7 protons on target. Both in Xsuite and BDSIM, the hadron decay to muons was biased by a factor of 100 to reduce the v...

    2025

  8. [8]

    The test was performed with the 500 mm long beryllium target and with all the colli- mators open to ±40 mm

    Results The two new optics configurations were tested during commissioning in April 2025 and dedicated tests in Oc- tober 2025, to be compared with the nominal optics for validation of the models. The test was performed with the 500 mm long beryllium target and with all the colli- mators open to ±40 mm. Figure 7 shows the measured electron rates normalize...

    2025

  9. [9]

    4), carrying between 57 % and 100 % of the parent hadron momentum, making the parametrization of an initial muon distribution a non- trivial task

    Fast muon phase-space modeling in M2 The muons in M2 are captured along the ∼700 m-long hadron decay section (see Fig. 4), carrying between 57 % and 100 % of the parent hadron momentum, making the parametrization of an initial muon distribution a non- trivial task. A more natural approach consists in modeling the production of the muons from the decays of...

  10. [10]

    The muon rate was mea- sured with an ionization chamber located at s = 1093 m from T6

    Results The optimized M2 muon optics were tested during a dedicated commissioning run in 2025, performed with the 500 mm long beryllium target in T6 and with all non- magnetic collimators fully open. The muon rate was mea- sured with an ionization chamber located at s = 1093 m from T6. The muon rate per proton on target was mea- sured for the nominal and ...

    2025

  11. [11]

    Banerjeeet al.,The North Experimental Area at the CERN Super Proton Synchrotron, Tech

    D. Banerjeeet al.,The North Experimental Area at the CERN Super Proton Synchrotron, Tech. Rep. (2021) ded- icated to Giorgio Brianti on the 50th anniversary of his founding the SPS Experimental Areas Group of CERN- Lab II and hence initiating the present Enterprise

    2021

  12. [12]

    Bernhardet al.,Report from the Conventional Beams Working Group to the Physics Beyond Colliders Study and to the 2026 European Strategy for Particle Physics Update, Tech

    J. Bernhardet al.,Report from the Conventional Beams Working Group to the Physics Beyond Colliders Study and to the 2026 European Strategy for Particle Physics Update, Tech. Rep. (CERN, Geneva. Geneva, 2025)

    2026

  13. [13]

    Fern´ andezet al.(PBC),Summary Report of the Physics Beyond Colliders Study at CERN, Tech

    A. Fern´ andezet al.(PBC),Summary Report of the Physics Beyond Colliders Study at CERN, Tech. Rep. (CERN, Geneva, 2025) arXiv:2505.00947

    Pith/arXiv arXiv 2025

  14. [14]

    Abbiendiet al.,Letter of Intent: the MUonE project, Tech

    G. Abbiendiet al.,Letter of Intent: the MUonE project, Tech. Rep. CERN-SPSC-2019-026, SPSC-I-252 (CERN, 2019)

    2019

  15. [15]

    M. N. Omidvar, X. Li, and X. Yao, IEEE Transactions on Evolutionary Computation26, 802 (2022)

    2022

  16. [16]

    M. N. Omidvar, X. Li, and X. Yao, IEEE Transactions on Evolutionary Computation26, 823 (2022)

    2022

  17. [17]

    Y. M. Andreevet al.(NA64), Phys. Rev. Lett.131, 161801 (2023), arXiv:2307.02404

    Pith/arXiv arXiv 2023

  18. [18]

    Y. M. Andreevet al.(NA64), Phys. Rev. Lett.133, 121803 (2024), arXiv:2406.03367

    Pith/arXiv arXiv 2024

  19. [19]

    Y. M. Andreevet al.(NA64), Phys. Rev. Lett.132, 211803 (2024), arXiv:2401.01708

    Pith/arXiv arXiv 2024

  20. [20]

    Adamset al.,COMPASS++/AMBER: Proposal for Measurements at the M2 beam line of the CERN SPS Phase-1: 2022-2024, Tech

    B. Adamset al.,COMPASS++/AMBER: Proposal for Measurements at the M2 beam line of the CERN SPS Phase-1: 2022-2024, Tech. Rep. (CERN, Geneva, 2019) the collaboration has not yet constituted itself, thus in- stead of a Spokesperson currently the nominated Contact Person is acting in place

    2022

  21. [21]

    de Maria and collaborators, Xsuite: Accelerator mod- eling and tracking toolkit (2022), software package

    R. de Maria and collaborators, Xsuite: Accelerator mod- eling and tracking toolkit (2022), software package. URL: https://xsuite.readthedocs.io/

    2022

  22. [22]

    Nevayet al., Computer Physics Communications252, 107200 (2020)

    L. Nevayet al., Computer Physics Communications252, 107200 (2020)

    2020

  23. [23]

    BDSIMversion 1.7.7, built on Geant4 [ 30–32] 10.7.2.3 with theFTFP BERTphysics list

  24. [24]

    Walker, A

    S. Walker, A. Abramov, L. Nevay, W. Shields, and S. Boogert, Computer Physics Communications272, 108228 (2022)

    2022

  25. [25]

    K. L. Brown, Adv. Part. Phys.1, 71 (1968)

    1968

  26. [26]

    Andreevet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrome- ters, Detectors and Associated Equipment1057, 168776 (2023)

    Y. Andreevet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrome- ters, Detectors and Associated Equipment1057, 168776 (2023)

    2023

  27. [27]

    Coet and N

    P. Coet and N. T. Doble,An introduction to the design of high-energy charged particle beams, Report (CERN, Geneva, 1986)

    1986

  28. [28]

    Gatignon, Design and tuning of secondary beamlines in the CERN north and east areas, CERN-ACC-NOTE- 2020-0043, updated 2025 (2025)

    L. Gatignon, Design and tuning of secondary beamlines in the CERN north and east areas, CERN-ACC-NOTE- 2020-0043, updated 2025 (2025)

    2020

  29. [29]

    Doble, L

    N. Doble, L. Gatignon, G. von Holtey, and F. Novoskolt- sev, Nuclear Instruments and Methods in Physics Re- search Section A: Accelerators, Spectrometers, Detectors and Associated Equipment343, 351 (1994)

    1994

  30. [30]

    Metzger,Studies for current and future high-intensity operation of CERN’s secondary M2 beamline, Ph.D

    F. Metzger,Studies for current and future high-intensity operation of CERN’s secondary M2 beamline, Ph.D. thesis, Rheinische Friedrich-Wilhelms-Universit¨ at Bonn (2024)

    2024

  31. [31]

    In H4, such a clear separation does not exist and Q13 was chosen in order to have one more degree of freedom than the number of figures of merit

    In M2, Q21 represents a natural breaking point, as it sep- arates the upstream hadron section and the downstream muon section. In H4, such a clear separation does not exist and Q13 was chosen in order to have one more degree of freedom than the number of figures of merit

  32. [32]

    K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, IEEE Trans. Evol. Comput.6, 182 (2002)

    2002

  33. [33]

    Saravanan, S

    R. Saravanan, S. Ramabalan, and C. Balamurugan, The International Journal of Advanced Manufacturing Tech- nology38, 1213 (2008)

    2008

  34. [34]

    Bergstra, R

    J. Bergstra, R. Bardenet, Y. Bengio, and B. K´ egl, Al- 14 gorithms for hyper-parameter optimization, Advances in Neural Information Processing Systems (NeurIPS) (2011)

    2011

  35. [35]

    Contributors, NSGAIIWithTPEWarmupSampler: Warm-started NSGA-II via TPE, OptunaHub (2024)

    O. Contributors, NSGAIIWithTPEWarmupSampler: Warm-started NSGA-II via TPE, OptunaHub (2024)

    2024

  36. [36]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, in Proceedings of the 25th ACM SIGKDD International Con- ference on Knowledge Discovery & Data Mining(2019) arXiv:1907.10902

    Pith/arXiv arXiv 2019

  37. [37]

    This is due to a combination of the so-calledcurse of dimensionality, so to say the exponential increase of the parameter space with the dimensionality of the problem, and to the fact that beam transmission is a sparse function. Intuitively, for any additional parameter with allowed po- larity change, the number of local maxima in transmission doubles, in...

  38. [38]

    D. N. Joanes and C. A. Gill,Comparing measures of sample skewness and kurtosis, Vol. 47 (1998) pp. 183–189, fisher-Pearson standardized moment coefficient

    1998

  39. [39]

    Garc ´ ıa-Pareja, A

    S. Garc ´ ıa-Pareja, A. M. Lallena, and F. Salvat, Frontiers in PhysicsV olume 9 - 2021, 10.3389/fphy.2021.718873 (2021)

    arXiv 2021

  40. [40]

    Allisonet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment835, 186 (2016)

    J. Allisonet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment835, 186 (2016)

    2016

  41. [41]

    Allisonet al., IEEE Transactions on Nuclear Science 53, 270 (2006)

    J. Allisonet al., IEEE Transactions on Nuclear Science 53, 270 (2006)

    2006

  42. [42]

    Agostinelliet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment506, 250 (2003)

    S. Agostinelliet al., Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment506, 250 (2003)

    2003

This paper was first reviewed by grok-4.5 on July 31, 2026.