Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints

A. Lemasson (1) ((1) GANIL; Bd Henri Becquerel; BP 55027; Caen Cedex 5; CEA/DRF - CNRS/IN2P3; F-14076; France); M. Rejmund (1)

arxiv: 2606.29466 · v1 · pith:4E3ESFN7new · submitted 2026-06-28 · 💻 cs.LG · nucl-ex· physics.ins-det

Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints

M. Rejmund (1) , A. Lemasson (1) ((1) GANIL , CEA/DRF - CNRS/IN2P3 , Bd Henri Becquerel , BP 55027 , F-14076 , Caen Cedex 5 , France) This is my paper

Pith reviewed 2026-06-30 07:52 UTC · model grok-4.3

classification 💻 cs.LG nucl-exphysics.ins-det

keywords self-supervised learninginstrument calibrationphysics-informed machine learningionic charge statepseudo-labellingVAMOS spectrometerdetector monitoring

0 comments

The pith

A physics-informed self-supervised framework jointly learns detector calibration parameters and ionic charge-state predictions directly from raw measurements without external labels.

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

The paper presents a method that transforms instrument calibration into a self-supervised optimization problem by using known physical constraints to generate pseudo-labels iteratively. It is demonstrated on ionic charge-state determination in the VAMOS++ spectrometer, where the calibration of a segmented ionization chamber and the inference of charge states are learned simultaneously starting from only a weak prior on the mean ionic charge state. The discrete nature of atomic masses drives progressive refinement through fractional pseudo-labelling. The resulting calibration coefficients provide a compact representation of detector state for monitoring drifts and aging. This establishes a general approach for self-calibrating scientific instruments without dedicated procedures or manually labelled data.

Core claim

The framework jointly learns latent detector calibration parameters and task-specific predictions directly from raw measurements by exploiting physical constraints to generate pseudo-labels iteratively, demonstrated on ionic charge-state determination where calibration and inference are learned simultaneously from a weak prior on mean ionic charge state, with the discrete nature of atomic masses supplying the refinement mechanism.

What carries the argument

Iterative fractional pseudo-labelling driven by the discrete nature of atomic masses, which generates pseudo-labels to jointly optimize calibration coefficients and charge-state predictions.

If this is right

Accurate ionic charge-state reconstruction is achieved without pre-calibrated signals.
Inferred calibration coefficients enable automated monitoring of gain drifts, pressure variations, and detector aging.
The resulting labels can be transferred to specialized models for quantifying detector imperfections and their evolution.
The approach provides a general paradigm for self-calibrating and self-monitoring scientific instruments.

Where Pith is reading between the lines

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

The same pseudo-labelling loop could extend to other detectors where physical quantities like energy levels or particle identities are discrete.
Real-time deployment might allow experiments to adjust acquisition parameters autonomously during data taking.
If the method generalizes, it could reduce reliance on periodic manual recalibrations in long-running facilities.

Load-bearing premise

The discrete nature of atomic masses together with other known physical constraints supplies sufficiently reliable pseudo-labels during iterative fractional labelling to bootstrap accurate calibration and predictions from only a weak prior on mean charge state.

What would settle it

Running the method on VAMOS++ raw data with the weak prior and finding that the learned charge-state predictions deviate from independently verified values or that calibration coefficients fail to track known detector drifts would falsify the claim.

Figures

Figures reproduced from arXiv: 2606.29466 by A. Lemasson (1) ((1) GANIL, Bd Henri Becquerel, BP 55027, Caen Cedex 5, CEA/DRF - CNRS/IN2P3, F-14076, France), M. Rejmund (1).

**Figure 1.** Figure 1: Physics-informed self-supervised calibration workflow: Weak physical priors initialize the model, after which iterative fractional pseudo-labelling driven by physical consistency constraints jointly refines the detector calibration coefficients and ionic charge-state predictions. The inferred calibration parameters can subsequently be monitored to characterize detector performance. 6. Experimental Evalua… view at source ↗

**Figure 2.** Figure 2: Training convergence: (a) The evolution of the RMSD(𝑞) for the training and validation, for initialization constraint ̄𝑞 = 39, obtained for the epoch before fractional pseudo-labelling and RMSD(𝐴) obtained during the pseudo-labelling as a function of the epoch number. The dashed ( ̄𝑞 = 36) and dotted ( ̄𝑞 = 36) lines indicate the training convergence of the model initialized with ̄𝑞 = 36 and ̄𝑞 = 42, respe… view at source ↗

**Figure 3.** Figure 3: Training progression: Evolution of the ionic charge-state and atomic mass spectra (a) and (b) pre-training, (c) and (d) convergence. These results demonstrate that the proposed framework successfully transforms physical consistency constraints into an effective source of supervision. Starting from raw detector signals and a weak prior on the average charge state, the model autonomously discovers the calibr… view at source ↗

**Figure 4.** Figure 4: Comparison with specialized network: The spectrum of the ionic charge state (𝑞) obtained using the calibration coefficients (𝑐𝑘 ) obtained with the proposed method. It is compared to the spectrum (𝑞𝑁𝑁 ) obtained using specialized network that accounts for the detection imperfections (Rejmund and Lemasson [12]) [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 6.** Figure 6: Atomic number identification chart: Twodimensional energy correlation plot between the energy loss (Δ𝐸 = ∑4 𝑘=0 𝑐𝑘𝐸 𝑟𝑎𝑤 𝑘 ) and residual energy (Δ𝐸𝑟𝑒𝑠 = ∑9 𝑘=5 𝑐𝑘𝐸 𝑟𝑎𝑤 𝑘 ) using the calibration coefficients (𝑐𝑘 ) obtained with the proposed method. The correlated bands correspond to different atomic numbers. generation of high-fidelity labels for atomic number (𝑍) identification, completing a fully self-su… view at source ↗

read the original abstract

Calibration remains one of the principal obstacles to the deployment of machine learning in scientific instrumentation because it typically relies on expert intervention, dedicated procedures, and manually labelled data. We introduce a physics-informed self-supervised framework that jointly learns latent detector calibration parameters and task-specific predictions directly from raw measurements without requiring pre-calibrated signals or external labels. The method exploits known physical constraints to generate pseudo-labels iteratively, transforming calibration into a self-supervised optimization problem. The approach is demonstrated for ionic charge-state determination in the VAMOS++ magnetic spectrometer, where the calibration of a segmented ionization chamber and the inference of ionic charge states are learned simultaneously. Starting from a weak prior on the mean ionic charge state, the model progressively refines its predictions through iterative fractional pseudo-labelling driven by the discrete nature of atomic masses. Beyond accurate ionic charge-state reconstruction, the inferred calibration coefficients provide a compact representation of the detector state that enables automated monitoring of gain drifts, pressure variations, and detector aging. The resulting labels can subsequently be transferred to specialized models that quantify detector imperfections and track their spatial and temporal evolution. These results establish a general paradigm for self-calibrating and self-monitoring scientific instruments and represent a step toward intelligent experimental systems capable of autonomous calibration, analysis, and performance optimization.

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

The paper outlines a self-supervised joint calibration and prediction method using physical constraints for pseudo-labeling on VAMOS++ data, but the abstract leaves the risk of degenerate solutions unaddressed.

read the letter

The main point is a self-supervised framework that learns latent detector calibration parameters together with predictions like ionic charge states, using known physical constraints to generate iterative pseudo-labels from raw measurements and a weak prior on mean charge state.

The work does a solid job identifying calibration as a practical bottleneck for ML in instrumentation and showing how discrete atomic masses plus other constraints can drive refinement without external labels. The side benefit of using the inferred coefficients for monitoring gain drifts and aging is a practical touch. The joint optimization and fractional pseudo-labelling on this spectrometer setup looks new relative to the cited absence of prior work.

The soft spot is the untested claim that the constraints break all degeneracies. The stress-test concern about multiple internally consistent but incorrect calibration solutions is reasonable given only the abstract; nothing shown rules out convergence to a wrong fixed point that still satisfies the weak prior. Without methods details, error bars, or comparisons to ground-truth calibrations, it is impossible to tell whether the procedure recovers accurate values or simply fits the supplied physics loosely.

This paper is for instrument scientists and physicists applying ML to detectors who want to reduce manual labeling. Readers working on physics-informed self-supervision would extract the core idea even if the VAMOS++ numbers need checking.

It deserves a serious referee to evaluate the actual derivations, convergence behavior, and validation experiments.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces a physics-informed self-supervised framework that jointly learns latent detector calibration parameters and task-specific predictions directly from raw measurements by iteratively generating pseudo-labels from physical consistency constraints. It is demonstrated on ionic charge-state determination in the VAMOS++ magnetic spectrometer, starting from a weak prior on the mean ionic charge state and using the discrete nature of atomic masses to refine predictions through iterative fractional pseudo-labelling; the inferred calibration coefficients are also positioned as a compact representation for automated monitoring of gain drifts, pressure variations, and detector aging.

Significance. If the central claim holds, the work would offer a meaningful advance toward autonomous scientific instrumentation by reducing reliance on expert calibration procedures and external labels. The byproduct of using calibration coefficients for detector-state monitoring is a practical strength. The approach could generalize to other instruments where physical constraints (e.g., discreteness of atomic masses) are available, provided the constraints prove sufficiently informative to resolve degeneracies.

major comments (2)

[Methods (iterative fractional pseudo-labelling)] The description of the iterative pseudo-labelling procedure (Methods section on the optimization loop) does not include a convergence analysis, uniqueness proof, or controlled experiment demonstrating that the physical constraints break all degeneracies and recover the true calibration coefficients rather than any of several self-consistent solutions that satisfy the weak mean-charge prior. This is load-bearing for the claim that calibration and inference are learned simultaneously without external validation data.
[Results (VAMOS++ demonstration)] Results on VAMOS++ report successful charge-state reconstruction but provide no quantitative comparison of the learned calibration coefficients against independently measured ground-truth values or ablation studies that disable individual physical constraints; without such checks it remains unclear whether the method recovers accurate parameters or merely fits the supplied prior.

minor comments (2)

[Abstract and Methods] The term 'fractional pseudo-labelling' is introduced in the abstract and methods without a precise definition or pseudocode; a short clarifying paragraph or algorithm box would improve reproducibility.
[Notation throughout] Notation for the latent calibration parameters and the weak prior should be introduced once with explicit symbols and reused consistently; occasional shifts between descriptive and symbolic forms reduce clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and for highlighting areas where additional rigor would strengthen the manuscript. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

Referee: [Methods (iterative fractional pseudo-labelling)] The description of the iterative pseudo-labelling procedure (Methods section on the optimization loop) does not include a convergence analysis, uniqueness proof, or controlled experiment demonstrating that the physical constraints break all degeneracies and recover the true calibration coefficients rather than any of several self-consistent solutions that satisfy the weak mean-charge prior. This is load-bearing for the claim that calibration and inference are learned simultaneously without external validation data.

Authors: We agree that a formal analysis of convergence and degeneracy resolution is important for the central claim. The current manuscript emphasizes the empirical demonstration on VAMOS++ data. In the revision we will add: (i) an empirical convergence study across random initializations and different weak priors, (ii) a short theoretical argument showing how the discrete atomic-mass constraint, combined with the mean-charge prior, eliminates the principal degeneracy classes, and (iii) a controlled synthetic-data experiment in which ground-truth calibration coefficients are known a priori and recovery is quantified. These additions will appear as a new subsection in Methods. revision: yes
Referee: [Results (VAMOS++ demonstration)] Results on VAMOS++ report successful charge-state reconstruction but provide no quantitative comparison of the learned calibration coefficients against independently measured ground-truth values or ablation studies that disable individual physical constraints; without such checks it remains unclear whether the method recovers accurate parameters or merely fits the supplied prior.

Authors: We acknowledge the absence of direct ground-truth calibration measurements for the VAMOS++ runs used; such independent calibrations are not routinely performed and are precisely the quantities the method aims to infer without external labels. We will therefore add ablation experiments that successively disable the mass-discreteness constraint and the mean-charge prior, reporting the resulting degradation in charge-state accuracy and calibration stability. We will also include indirect validation by showing that the inferred coefficients produce physically plausible detector-state trajectories consistent with known gain-drift and pressure effects. These changes will be incorporated into the Results section. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external physical constraints rather than self-definition or fitted inputs

full rationale

The abstract describes an iterative pseudo-labelling process that starts from a weak prior on mean charge state and refines predictions using the discrete nature of atomic masses and other physical constraints. No equations, self-citations, or ansatzes are provided that would reduce the output predictions or calibration parameters to the input prior by construction. The method claims to exploit known external physical facts (atomic mass discreteness) to generate pseudo-labels, which constitutes independent content rather than a renaming or self-referential fit. Without load-bearing self-citations or explicit reduction of the learned quantities to the supplied prior, the framework is self-contained against external benchmarks and receives a zero circularity score.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review; ledger populated from stated elements in the abstract.

free parameters (1)

weak prior on mean ionic charge state
Explicitly stated starting point used to initiate iterative refinement; value not given but treated as input.

axioms (1)

domain assumption Known physical constraints including the discrete nature of atomic masses generate reliable pseudo-labels
Invoked to drive iterative fractional pseudo-labelling without external labels.

pith-pipeline@v0.9.1-grok · 5801 in / 1261 out tokens · 23826 ms · 2026-06-30T07:52:29.037060+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

29 extracted references · 23 canonical work pages

[1]

Performance of VAMOS for reactions near the Coulomb Page 10 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints barrier

Pullanhiotan,S.,Rejmund,M.,Navin,A.,Mittig,W.,Bhattacharyya, S., 2008. Performance of VAMOS for reactions near the Coulomb Page 10 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints barrier. Nuclear Instruments and Methods in Physics Research Sec- tion A 593, 343–352. URL:http://www.sciencedirect.com/science...

work page doi:10.1016/j.nima.2008.05.003 2008
[2]

Perfor- mance of the improved larger acceptance spectrometer: VAMOS++

Rejmund, M., Lecornu, B., Navin, A., Schmitt, C., Damoy, S., De- laune, O., Enguerrand, J., Fremont, G., Gangnant, P., Gaudefroy, L., Jacquot, B., Pancin, J., Pullanhiotan, S., Spitaels, C., 2011. Perfor- mance of the improved larger acceptance spectrometer: VAMOS++. NuclearInstrumentsandMethodsinPhysicsResearchSectionA646, 184–191. URL:http://www.science...

work page doi:10.1016/j.nima.2011.05.007 2011
[3]

Machine learning at the energy and intensity frontiers of particle physics

Radovic, A., Williams, M., Rousseau, D., Kagan, M., Bonacorsi, D., Himmel, A., Aurisano, A., Terao, K., Wongjirad, T., 2018. Machine learning at the energy and intensity frontiers of particle physics. Na- ture 560, 41–48. URL:https://doi.org/10.1038/s41586-018-0361-2, doi:10.1038/s41586-018-0361-2

work page doi:10.1038/s41586-018-0361-2 2018
[4]

Deep learning and its application to lhc physics

Guest, D., Cranmer, K., Whiteson, D., 2018. Deep learning and its application to lhc physics. Annual Review of Nuclear and Particle Science68,161–181. URL:https://www.annualreviews.org/content/ journals/10.1146/annurev-nucl-101917-021019, doi:https://doi.org/ 10.1146/annurev-nucl-101917-021019

work page doi:10.1146/annurev-nucl-101917-021019 2018
[5]

Machine learning and the physical sciences

Carleo,G.,Cirac,I.,Cranmer,K.,Daudet,L.,Schuld,M.,Tishby,N., Vogt-Maranto, L., Zdeborová, L., 2019. Machine learning and the physical sciences. Rev. Mod. Phys. 91, 045002. URL:https://link. aps.org/doi/10.1103/RevModPhys.91.045002, doi:10.1103/RevModPhys. 91.045002

work page doi:10.1103/revmodphys.91.045002 2019
[6]

Machine learning in high energy physics community white paper

Albertsson, K., Altoe, P., Anderson, D., Andrews, M., Araque Es- pinosa, J.P., Aurisano, A., Basara, L., Bevan, A., Bhimji, W., Bona- corsi, D., Calafiura, P., Campanelli, M., Capps, L., Carminati, F., Carrazza, S., Childers, T., Coniavitis, E., Cranmer, K., David, C., Davis, D., Duarte, J., Erdmann, M., Eschle, J., Farbin, A., Feickert, M., Castro, N.F.,...

work page doi:10.1088/1742-6596/1085/2/022008 2018
[7]

Pseudo-label : The simple and efficient semi- supervised learning method for deep neural networks

Lee, D.H., 2013. Pseudo-label : The simple and efficient semi- supervised learning method for deep neural networks. URL:https: //api.semanticscholar.org/CorpusID:18507866

2013
[8]

nuScenes: a multimodal dataset for autonomous driving

Xie, Q., Luong, M.T., Hovy, E., Le, Q.V., 2020. Self-training with noisy student improves imagenet classification, in: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10684–10695. doi:10.1109/CVPR42600.2020.01070

work page doi:10.1109/cvpr42600.2020.01070 2020
[9]

Fixmatch: Simplifying semi-supervised learning with consistency and confidence, in: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H

Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C., Cubuk, E.D., Kurakin, A., Li, C.L., 2020. Fixmatch: Simplifying semi-supervised learning with consistency and confidence, in: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H. (Eds.), Advances in Neural Information Processing Systems, Curran Associates, Inc.. pp. 596–608. ...

2020
[10]

A survey on semi-supervised learning

van Engelen, J.E., Hoos, H.H., 2020. A survey on semi-supervised learning. Machine Learning 109, 373–440. URL:https://doi.org/ 10.1007/s10994-019-05855-6, doi:10.1007/s10994-019-05855-6

work page doi:10.1007/s10994-019-05855-6 2020
[11]

Maximum likelihood from incomplete data via the EM algorithm

Dempster,A.P.,Laird,N.M.,Rubin,D.B.,1977.Maximumlikelihood from incomplete data via the em algorithm. Journal of the Royal StatisticalSociety:SeriesB(Methodological)39,1–22. doi:10.1111/ j.2517-6161.1977.tb01600.x

work page arXiv 1977
[12]

Analysis of atomic charge state andatomicnumberforVAMOS++magneticspectrometerusingdeep neural networks and fractionally labelled events

Rejmund, M., Lemasson, A., 2025. Analysis of atomic charge state andatomicnumberforVAMOS++magneticspectrometerusingdeep neural networks and fractionally labelled events. Journal of Instru- mentation20,P08022. URL:https://dx.doi.org/10.1088/1748-0221/ 20/08/P08022, doi:10.1088/1748-0221/20/08/P08022

work page doi:10.1088/1748-0221/ 2025
[13]

Physics- informed neural networks: A deep learning framework for solv- ing forward and inverse problems involving nonlinear partial dif- ferential equations

Raissi, M., Perdikaris, P., Karniadakis, G., 2019. Physics- informed neural networks: A deep learning framework for solv- ing forward and inverse problems involving nonlinear partial dif- ferential equations. Journal of Computational Physics 378, 686–

2019
[14]

URL:https://www.sciencedirect.com/science/article/pii/ S0021999118307125, doi:https://doi.org/10.1016/j.jcp.2018.10.045

work page doi:10.1016/j.jcp.2018.10.045 2018
[15]

Physics-informed machine learning,

Karniadakis, G.E., Kevrekidis, I.G., Lu, L., Perdikaris, P., Wang, S., Yang, L., 2021. Physics-informed machine learning. Na- ture Reviews Physics 3, 422–440. URL:https://doi.org/10.1038/ s42254-021-00314-5, doi:10.1038/s42254-021-00314-5

work page doi:10.1038/s42254-021-00314-5 2021
[16]

Scientific

Cuomo, S., Di Cola, V.S., Giampaolo, F., Rozza, G., Raissi, M., Piccialli, F., 2022. Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s Next. Journal of Scientific Computing 92, 88. URL:https://doi.org/10. 1007/s10915-022-01939-z, doi:10.1007/s10915-022-01939-z

work page doi:10.1007/s10915-022-01939-z 2022
[17]

Willard, X

Willard, J., Jia, X., Xu, S., Steinbach, M., Kumar, V., 2022. Inte- grating scientific knowledge with machine learning for engineering andenvironmentalsystems.URL:https://arxiv.org/abs/2003.04919, arXiv:2003.04919

work page arXiv 2022
[18]

Data-drivenscienceandengineering

Brunton,S.L.,Kutz,J.N.,2022. Data-drivenscienceandengineering. Cambridge University Press eBooks doi:10.1017/9781009089517

work page doi:10.1017/9781009089517 2022
[19]

Accelerating the discovery of materials for cleanenergyintheeraofsmartautomation.NatureReviewsMaterials 3, 5–20

Tabor, D.P., Roch, L.M., Saikin, S.K., Kreisbeck, C., Sheberla, D., Montoya, J.H., Dwaraknath, S., Aykol, M., Ortiz, C., Tribukait, H., Amador-Bedolla, C., Brabec, C.J., Maruyama, B., Persson, K.A., Aspuru-Guzik, A., 2018. Accelerating the discovery of materials for cleanenergyintheeraofsmartautomation.NatureReviewsMaterials 3, 5–20. URL:https://doi.org/1...

work page doi:10.1038/s41578-018-0005-z 2018
[20]

Autonomous experimentation systems for materials development: A community perspective

Stach, E., DeCost, B., Kusne, A.G., Hattrick-Simpers, J., Brown, K.A., Reyes, K.G., Schrier, J., Billinge, S., Buonassisi, T., Foster, I., Gomes, C.P., Gregoire, J.M., Mehta, A., Montoya, J., Olivetti, E., Park, C., Rotenberg, E., Saikin, S.K., Smullin, S., Stanev, V., Maruyama, B., 2021. Autonomous experimentation systems for materials development: A com...

2021
[21]

URL:https://www.sciencedirect.com/science/article/pii/ S2590238521003064, doi:https://doi.org/10.1016/j.matt.2021.06.036

work page doi:10.1016/j.matt.2021.06.036 2021
[22]

Autonomous experiments using active learning and AI

Ren,Z.,Ren,Z.,Zhang,Z.,Buonassisi,T.,Li,J.,2023. Autonomous experiments using active learning and AI. Nature Reviews Materi- als 8, 563–564. URL:https://doi.org/10.1038/s41578-023-00588-4, doi:10.1038/s41578-023-00588-4

work page doi:10.1038/s41578-023-00588-4 2023
[23]

Digital twin: Values, challenges and enablers from a modeling perspective

Rasheed, A., San, O., Kvamsdal, T., 2020. Digital twin: Values, challenges and enablers from a modeling perspective. IEEE Access 8, 21980–22012. doi:10.1109/ACCESS.2020.2970143

work page doi:10.1109/access.2020.2970143 2020
[24]

IEEE Transactions on Nuclear Science 63, 878–897

Edelen, A.L., Biedron, S.G., Chase, B.E., Edstrom, D., Milton, S.V., Stabile,P.,2016.Neuralnetworksformodelingandcontrolofparticle accelerators. IEEE Transactions on Nuclear Science 63, 878–897. doi:10.1109/TNS.2016.2543203

work page doi:10.1109/tns.2016.2543203 2016
[25]

Machine learning-based longitudinal phase space prediction of particle accelerators

Emma, C., Edelen, A., Hogan, M.J., O’Shea, B., White, G., Yakimenko, V., 2018. Machine learning-based longitudinal phase space prediction of particle accelerators. Phys. Rev. Accel. Beams 21, 112802. URL:https://link.aps.org/doi/10.1103/ PhysRevAccelBeams.21.112802, doi:10.1103/PhysRevAccelBeams.21. 112802

work page doi:10.1103/physrevaccelbeams.21 2018
[26]

Machine learning for design and controlofparticleaccelerators:Alookbackwardandforward.Annual Review of Nuclear and Particle Science 74, 557–581

Edelen, A., Huang, X., 2024. Machine learning for design and controlofparticleaccelerators:Alookbackwardandforward.Annual Review of Nuclear and Particle Science 74, 557–581. doi:10.1146/ annurev-nucl-121423-100719. Page 11 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints

2024
[27]

Bayesian optimization algorithms for accelerator physics

Roussel, R., et al., 2024. Bayesian optimization algorithms for accelerator physics. Phys. Rev. Accel. Beams 27, 084801. doi:10. 1103/PhysRevAccelBeams.27.084801,arXiv:2312.05667

work page arXiv 2024
[28]

Seven-dimensional trajectory reconstruction for VAMOS++

Rejmund, M., Lemasson, A., 2025. Seven-dimensional trajectory reconstruction for VAMOS++. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1076, 170445. URL:https://www. sciencedirect.com/science/article/pii/S0168900225002463https: //linkinghub.elsevier.com/retrieve/pii/S0168900...

work page doi:10.1016/j.nima.2025.170445 2025
[29]

Page 12 of 12

E826-21Collaboration,2022.E826-21GANILdataset.doi:10.26143/ GANIL-2022-E826_21. Page 12 of 12

2022

[1] [1]

Performance of VAMOS for reactions near the Coulomb Page 10 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints barrier

Pullanhiotan,S.,Rejmund,M.,Navin,A.,Mittig,W.,Bhattacharyya, S., 2008. Performance of VAMOS for reactions near the Coulomb Page 10 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints barrier. Nuclear Instruments and Methods in Physics Research Sec- tion A 593, 343–352. URL:http://www.sciencedirect.com/science...

work page doi:10.1016/j.nima.2008.05.003 2008

[2] [2]

Perfor- mance of the improved larger acceptance spectrometer: VAMOS++

Rejmund, M., Lecornu, B., Navin, A., Schmitt, C., Damoy, S., De- laune, O., Enguerrand, J., Fremont, G., Gangnant, P., Gaudefroy, L., Jacquot, B., Pancin, J., Pullanhiotan, S., Spitaels, C., 2011. Perfor- mance of the improved larger acceptance spectrometer: VAMOS++. NuclearInstrumentsandMethodsinPhysicsResearchSectionA646, 184–191. URL:http://www.science...

work page doi:10.1016/j.nima.2011.05.007 2011

[3] [3]

Machine learning at the energy and intensity frontiers of particle physics

Radovic, A., Williams, M., Rousseau, D., Kagan, M., Bonacorsi, D., Himmel, A., Aurisano, A., Terao, K., Wongjirad, T., 2018. Machine learning at the energy and intensity frontiers of particle physics. Na- ture 560, 41–48. URL:https://doi.org/10.1038/s41586-018-0361-2, doi:10.1038/s41586-018-0361-2

work page doi:10.1038/s41586-018-0361-2 2018

[4] [4]

Deep learning and its application to lhc physics

Guest, D., Cranmer, K., Whiteson, D., 2018. Deep learning and its application to lhc physics. Annual Review of Nuclear and Particle Science68,161–181. URL:https://www.annualreviews.org/content/ journals/10.1146/annurev-nucl-101917-021019, doi:https://doi.org/ 10.1146/annurev-nucl-101917-021019

work page doi:10.1146/annurev-nucl-101917-021019 2018

[5] [5]

Machine learning and the physical sciences

Carleo,G.,Cirac,I.,Cranmer,K.,Daudet,L.,Schuld,M.,Tishby,N., Vogt-Maranto, L., Zdeborová, L., 2019. Machine learning and the physical sciences. Rev. Mod. Phys. 91, 045002. URL:https://link. aps.org/doi/10.1103/RevModPhys.91.045002, doi:10.1103/RevModPhys. 91.045002

work page doi:10.1103/revmodphys.91.045002 2019

[6] [6]

Machine learning in high energy physics community white paper

Albertsson, K., Altoe, P., Anderson, D., Andrews, M., Araque Es- pinosa, J.P., Aurisano, A., Basara, L., Bevan, A., Bhimji, W., Bona- corsi, D., Calafiura, P., Campanelli, M., Capps, L., Carminati, F., Carrazza, S., Childers, T., Coniavitis, E., Cranmer, K., David, C., Davis, D., Duarte, J., Erdmann, M., Eschle, J., Farbin, A., Feickert, M., Castro, N.F.,...

work page doi:10.1088/1742-6596/1085/2/022008 2018

[7] [7]

Pseudo-label : The simple and efficient semi- supervised learning method for deep neural networks

Lee, D.H., 2013. Pseudo-label : The simple and efficient semi- supervised learning method for deep neural networks. URL:https: //api.semanticscholar.org/CorpusID:18507866

2013

[8] [8]

nuScenes: a multimodal dataset for autonomous driving

Xie, Q., Luong, M.T., Hovy, E., Le, Q.V., 2020. Self-training with noisy student improves imagenet classification, in: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10684–10695. doi:10.1109/CVPR42600.2020.01070

work page doi:10.1109/cvpr42600.2020.01070 2020

[9] [9]

Fixmatch: Simplifying semi-supervised learning with consistency and confidence, in: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H

Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C., Cubuk, E.D., Kurakin, A., Li, C.L., 2020. Fixmatch: Simplifying semi-supervised learning with consistency and confidence, in: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H. (Eds.), Advances in Neural Information Processing Systems, Curran Associates, Inc.. pp. 596–608. ...

2020

[10] [10]

A survey on semi-supervised learning

van Engelen, J.E., Hoos, H.H., 2020. A survey on semi-supervised learning. Machine Learning 109, 373–440. URL:https://doi.org/ 10.1007/s10994-019-05855-6, doi:10.1007/s10994-019-05855-6

work page doi:10.1007/s10994-019-05855-6 2020

[11] [11]

Maximum likelihood from incomplete data via the EM algorithm

Dempster,A.P.,Laird,N.M.,Rubin,D.B.,1977.Maximumlikelihood from incomplete data via the em algorithm. Journal of the Royal StatisticalSociety:SeriesB(Methodological)39,1–22. doi:10.1111/ j.2517-6161.1977.tb01600.x

work page arXiv 1977

[12] [12]

Analysis of atomic charge state andatomicnumberforVAMOS++magneticspectrometerusingdeep neural networks and fractionally labelled events

Rejmund, M., Lemasson, A., 2025. Analysis of atomic charge state andatomicnumberforVAMOS++magneticspectrometerusingdeep neural networks and fractionally labelled events. Journal of Instru- mentation20,P08022. URL:https://dx.doi.org/10.1088/1748-0221/ 20/08/P08022, doi:10.1088/1748-0221/20/08/P08022

work page doi:10.1088/1748-0221/ 2025

[13] [13]

Physics- informed neural networks: A deep learning framework for solv- ing forward and inverse problems involving nonlinear partial dif- ferential equations

Raissi, M., Perdikaris, P., Karniadakis, G., 2019. Physics- informed neural networks: A deep learning framework for solv- ing forward and inverse problems involving nonlinear partial dif- ferential equations. Journal of Computational Physics 378, 686–

2019

[14] [14]

URL:https://www.sciencedirect.com/science/article/pii/ S0021999118307125, doi:https://doi.org/10.1016/j.jcp.2018.10.045

work page doi:10.1016/j.jcp.2018.10.045 2018

[15] [15]

Physics-informed machine learning,

Karniadakis, G.E., Kevrekidis, I.G., Lu, L., Perdikaris, P., Wang, S., Yang, L., 2021. Physics-informed machine learning. Na- ture Reviews Physics 3, 422–440. URL:https://doi.org/10.1038/ s42254-021-00314-5, doi:10.1038/s42254-021-00314-5

work page doi:10.1038/s42254-021-00314-5 2021

[16] [16]

Scientific

Cuomo, S., Di Cola, V.S., Giampaolo, F., Rozza, G., Raissi, M., Piccialli, F., 2022. Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s Next. Journal of Scientific Computing 92, 88. URL:https://doi.org/10. 1007/s10915-022-01939-z, doi:10.1007/s10915-022-01939-z

work page doi:10.1007/s10915-022-01939-z 2022

[17] [17]

Willard, X

Willard, J., Jia, X., Xu, S., Steinbach, M., Kumar, V., 2022. Inte- grating scientific knowledge with machine learning for engineering andenvironmentalsystems.URL:https://arxiv.org/abs/2003.04919, arXiv:2003.04919

work page arXiv 2022

[18] [18]

Data-drivenscienceandengineering

Brunton,S.L.,Kutz,J.N.,2022. Data-drivenscienceandengineering. Cambridge University Press eBooks doi:10.1017/9781009089517

work page doi:10.1017/9781009089517 2022

[19] [19]

Accelerating the discovery of materials for cleanenergyintheeraofsmartautomation.NatureReviewsMaterials 3, 5–20

Tabor, D.P., Roch, L.M., Saikin, S.K., Kreisbeck, C., Sheberla, D., Montoya, J.H., Dwaraknath, S., Aykol, M., Ortiz, C., Tribukait, H., Amador-Bedolla, C., Brabec, C.J., Maruyama, B., Persson, K.A., Aspuru-Guzik, A., 2018. Accelerating the discovery of materials for cleanenergyintheeraofsmartautomation.NatureReviewsMaterials 3, 5–20. URL:https://doi.org/1...

work page doi:10.1038/s41578-018-0005-z 2018

[20] [20]

Autonomous experimentation systems for materials development: A community perspective

Stach, E., DeCost, B., Kusne, A.G., Hattrick-Simpers, J., Brown, K.A., Reyes, K.G., Schrier, J., Billinge, S., Buonassisi, T., Foster, I., Gomes, C.P., Gregoire, J.M., Mehta, A., Montoya, J., Olivetti, E., Park, C., Rotenberg, E., Saikin, S.K., Smullin, S., Stanev, V., Maruyama, B., 2021. Autonomous experimentation systems for materials development: A com...

2021

[21] [21]

URL:https://www.sciencedirect.com/science/article/pii/ S2590238521003064, doi:https://doi.org/10.1016/j.matt.2021.06.036

work page doi:10.1016/j.matt.2021.06.036 2021

[22] [22]

Autonomous experiments using active learning and AI

Ren,Z.,Ren,Z.,Zhang,Z.,Buonassisi,T.,Li,J.,2023. Autonomous experiments using active learning and AI. Nature Reviews Materi- als 8, 563–564. URL:https://doi.org/10.1038/s41578-023-00588-4, doi:10.1038/s41578-023-00588-4

work page doi:10.1038/s41578-023-00588-4 2023

[23] [23]

Digital twin: Values, challenges and enablers from a modeling perspective

Rasheed, A., San, O., Kvamsdal, T., 2020. Digital twin: Values, challenges and enablers from a modeling perspective. IEEE Access 8, 21980–22012. doi:10.1109/ACCESS.2020.2970143

work page doi:10.1109/access.2020.2970143 2020

[24] [24]

IEEE Transactions on Nuclear Science 63, 878–897

Edelen, A.L., Biedron, S.G., Chase, B.E., Edstrom, D., Milton, S.V., Stabile,P.,2016.Neuralnetworksformodelingandcontrolofparticle accelerators. IEEE Transactions on Nuclear Science 63, 878–897. doi:10.1109/TNS.2016.2543203

work page doi:10.1109/tns.2016.2543203 2016

[25] [25]

Machine learning-based longitudinal phase space prediction of particle accelerators

Emma, C., Edelen, A., Hogan, M.J., O’Shea, B., White, G., Yakimenko, V., 2018. Machine learning-based longitudinal phase space prediction of particle accelerators. Phys. Rev. Accel. Beams 21, 112802. URL:https://link.aps.org/doi/10.1103/ PhysRevAccelBeams.21.112802, doi:10.1103/PhysRevAccelBeams.21. 112802

work page doi:10.1103/physrevaccelbeams.21 2018

[26] [26]

Machine learning for design and controlofparticleaccelerators:Alookbackwardandforward.Annual Review of Nuclear and Particle Science 74, 557–581

Edelen, A., Huang, X., 2024. Machine learning for design and controlofparticleaccelerators:Alookbackwardandforward.Annual Review of Nuclear and Particle Science 74, 557–581. doi:10.1146/ annurev-nucl-121423-100719. Page 11 of 12 Self-Supervised Calibration of Scientific Instruments Using Physical Consistency Constraints

2024

[27] [27]

Bayesian optimization algorithms for accelerator physics

Roussel, R., et al., 2024. Bayesian optimization algorithms for accelerator physics. Phys. Rev. Accel. Beams 27, 084801. doi:10. 1103/PhysRevAccelBeams.27.084801,arXiv:2312.05667

work page arXiv 2024

[28] [28]

Seven-dimensional trajectory reconstruction for VAMOS++

Rejmund, M., Lemasson, A., 2025. Seven-dimensional trajectory reconstruction for VAMOS++. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1076, 170445. URL:https://www. sciencedirect.com/science/article/pii/S0168900225002463https: //linkinghub.elsevier.com/retrieve/pii/S0168900...

work page doi:10.1016/j.nima.2025.170445 2025

[29] [29]

Page 12 of 12

E826-21Collaboration,2022.E826-21GANILdataset.doi:10.26143/ GANIL-2022-E826_21. Page 12 of 12

2022