Mass Agnostic Jet Taggers

Andrea Mitridate; Bryan Ostdiek; Layne Bradshaw; Rashmish K. Mishra

REVIEW 3 major objections 6 minor 2 cited by

Mass Agnostic Jet Taggers

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Planing and PCA match adversarial taggers at ~100x less training time

desk verdict A competent and genuinely useful benchmark of mass-decorrelation taggers, but the headline equivalence between cheap and expensive methods rests on a single adversarial run without error bars, so the central comparison needs a revision before the claim is solid. read the letter →

arxiv 1908.08959 v2 pith:DRTCBNHW submitted 2019-08-23 hep-ph hep-ex

Layne Bradshaw , Rashmish K. Mishra , Andrea Mitridate , Bryan Ostdiek This is my paper · ORCID

classification hep-phhep-ex

keywords jetsubstructuremassdecorrelationtaggingdataplaningPCArescalingadversarialnetworksuBoostBhattacharyyadistance

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

Jet taggers that separate new-physics signals from ordinary QCD background tend to distort the background's invariant-mass shape, making bump-hunt searches harder. This paper compares four decorrelation strategies on 2-, 3-, and 4-prong jets and claims that the two data-augmentation approaches—Planing and PCA-based rescaling—match the two training-augmentation approaches—adversarial networks and uBoost—in preserving the background distribution, while being far simpler and roughly 100 times faster to train. The authors quantify sculpting with the Bhattacharyya distance between the original background mass distribution and the distribution after cuts, and find that a network trained on planed data comes close to the adversarial network for 3- and 4-prong jets, while PCA-based scaling parallels uBoost. If true, this gives experimental analyses a cheap way to gain most of the decorrelation benefit without the tuning overhead of adversarial training.

What carries the argument

The machinery has three parts. The input representation is the 5-body N-subjettiness basis, 11 variables $\tau^{(\beta)}_N$ from Eq. (4), chosen so no classifier sees the jet scale. The decorrelation procedures are (i) planing, with event weights $[w(X_i)]^{-1}=C\,d\sigma(X_i)/dm|_{m=m_i}$ that flatten the jet-mass distribution of each class, and (ii) PCA rescaling, $\vec\tau^{\rm PCA}_i = R_i^{-1} S_i R_i \vec\tau^{\rm std}_i$, which removes per-mass-bin linear correlations of the inputs. Against these are set uBoost, a boosted-decision-tree variant that up-weights events whose local efficiency deviates from the target, and the adversarial network, whose classifier minimizes $L_{\rm tagger}=L_{\rm classifier}-\lambda L_{\rm adversary}$ with the adversary predicting the mass bin. The comparison metric is the Bhattacharyya distance, Eq. (9), between the QCD mass distribution before and after applying a signal-efficiency cut.

What would settle it

Repeat the comparison with jet-image inputs instead of the 11 N-subjettiness variables and recompute Bhattacharyya distance at fixed background rejection; if planing and PCA no longer track adversarial and uBoost, the representation-invariance premise fails.

Watch

Extended reading notes

Core claim

At the paper's core is a quantitative comparison, on a single 11-variable N-subjettiness basis, of two families of mass-decorrelation methods for tagging 2-, 3-, and 4-prong jets against a QCD background. The claim is that the data-augmentation methods—planing, which reweights events so both signal and background have a uniform jet-mass distribution, and PCA rescaling, which per-mass-bin whitens and rotates the input features—reach approximately the same level of background-shape preservation as the training-augmentation methods—adversarial networks and uBoost—when measured by the Bhattacharyya distance between the background mass distribution before and after cuts. For 3- and 4-prong jets, a neural network trained on planed data tracks the adversarial network's sculpting curve nearly identically while training about 100 times faster; PCA-based scaling gives curves similar to uBoost while costing far less time. Because no analytic decorrelated tagger exists for 3- or 4-prong jets, these results position data augmentation as the practical route to mass-agnostic multivariate tagging.

Load-bearing premise

All classifiers were trained on the same 11-variable N-subjettiness basis, so the relative ranking of methods could differ if a different input representation changed how hard each decorrelation task is.

Editorial extensions

If this is right

Analyses can adopt planing or PCA rescaling as drop-in preprocessing without changing the classifier architecture or tuning a new loss-function hyperparameter.
For 3- and 4-prong signals, planed neural networks preserve the background shape almost as well as adversarial networks while cutting training time by a factor of about 100.
PCA rescaling paired with a boosted decision tree gives uBoost-like sculpting at less than 1/20 of the training time, enabling fast iteration during analysis development.
Since the taggers rely on substructure and not the absolute jet scale, a single planed or PCA-trained classifier may serve across a range of signal masses, avoiding mass-by-mass retraining.

Reading between the lines

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

If the method ordering is representation-independent, the same study on jet images or graph-based representations should reproduce the qualitative result; this is a direct experimental test of the paper's implicit assumption.
The near-agreement between planing and adversarial training suggests the adversary may be learning to flatten the same background density that planing flattens explicitly; comparing the implied event weights could reveal the two methods are solving the same optimization.
Extending planing to two protected variables, jet mass and transverse momentum, would likely yield a tagger robust to both dominant kinematic systematics; the paper lists multidimensional planing as future work.

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.

Desk Editor's Note

read the letter

What you should know: this paper benchmarks four mass-decorrelation methods—planing, PCA rescaling, uBoost, and adversarial training—on 2-, 3-, and 4-prong jets, and argues that the cheap data-augmentation methods (planing and PCA) match the expensive training-augmentation methods in preserving the QCD background shape. That is a useful practical message for the jet-substructure community, and the study is the first to do this systematically beyond the 2-body case that ATLAS covered.

The paper does several things well. The method descriptions are clear, the choice of metrics (Bhattacharyya distance, with Jensen-Shannon as a cross-check) is sensible, and the training-time comparison in Table 2 is genuinely informative. The extension to 3- and 4-prong topologies matters because the analytic DDT approach does not generalize, and the paper shows where the alternatives fill that gap. They also ship code, which makes the benchmark reproducible. Credit where due: this is a solid, well-organized study.

The soft spot is exactly where the reader and the stress-test point. The central claim that planing and adversarial training give “nearly identical” background distortion for 3- and 4-prong jets is built on Fig. 15, and the adversarial curve there appears to come from a single training run (the Table 2 footnote says adversarial statistics are one sample per λ value). The planed networks have 10-seed statistics; the adversarial one does not. Without error bars on that curve, the claimed equivalence could be run-to-run noise. The problem is worse at high background rejection, where the 4-prong background is sparse. This is a legitimate referee issue, not a manufactured one. I would also have liked the abstract to say “comparable performance” rather than “similar performance” until the uncertainty is quantified.

A second, milder limitation is that the whole comparison uses one input representation (the 11-variable N-subjettiness basis). The authors acknowledge this and leave images/sequences/graphs to future work. That is fine for a first systematic study, but it means the conclusion might not transfer to other representations. I do not see this as a fatal flaw—the paper is explicit about the scope.

Bottom line: the paper is worth engaging with seriously. The qualitative conclusion—planing and PCA are competitive with uBoost and adversarial training at a fraction of the cost—is probably correct, but the quantitative “nearly identical” claim needs error bars before I would trust it. Send it to peer review; with a moderate revision it would be a useful reference.

Referee Report

3 major / 6 minor

Summary. This paper presents a systematic comparison of methods for decorrelating jet taggers from the jet mass, applied to 2-, 3-, and 4-prong signal jets against a QCD background. The methods considered are single-variable taggers (τ21, τ32, τ43, and the analytic τDDT21), standard multivariate classifiers (BDT and NN), and four decorrelation approaches: PCA-based rescaling, Planing, uBoost, and adversarial neural networks. All multivariate classifiers use the same 11-variable N-subjettiness input basis of Eq. (4), and performance is evaluated with ROC curves, background-rejection curves, the Bhattacharyya distance between the background mass distribution before and after cuts, and measured training times. The main quantitative claim is that the data-augmentation methods Planing and PCA deliver performance comparable to the training-augmentation methods uBoost and adversarial networks, while requiring a fraction of the training time.

Significance. If the stated equivalence holds, the paper gives practically useful guidance for LHC analyses: cheap and simple data-augmentation methods can replace expensive adversarial training in several regimes, and the comparison extends beyond the 2-prong case studied by ATLAS. The study is well organized, uses multiple quantitative metrics, provides a parameter scan for the adversarial weight, and states that reproducing code is available on GitHub. The main weakness is that the headline 'similar performance' is not yet backed by uncertainty quantification: the adversarial network is represented by a single training run at the chosen λ, and none of the Bhattacharyya-distance curves carry error bars. The paper is therefore more convincing as a benchmark study than as a proof of equivalence between methods.

major comments (3)

[4.3, Fig. 15; Table 2 footnote] The central claim that Planing performs nearly identically to the adversarial network for 3- and 4-prong jets is not supported with uncertainty quantification. Table 2 states that the adversarial training statistics are 'over sampled once over each of the nine different values of λ due to the long training time'; unlike the base, PCA, and Planed networks, there are no 10-seed repeats for the λ=50 classifier shown in Fig. 15. The plotted separation between the planed and adversarial curves could therefore be smaller than run-to-run seed variation. Please add repeated adversarial training runs at the chosen λ (with at least the same number of seeds as the other methods) and show error bands on the Bhattacharyya-distance curves, or explicitly rephrase the 'nearly identical' claim as a single-run observation.
[4.1, Fig. 15] The 4-prong comparisons are affected by low statistics in the high-rejection tail. The text notes that for the 4-prong signal the background is exhausted near ε_S ≈ 0.1, so the Bhattacharyya distances at high background rejection are estimated from very few surviving QCD events. Without statistical error bars on dB, the apparent agreement between Planing and the adversarial network in the right-hand panels of Fig. 15 cannot be distinguished from noise. This is especially relevant because the abstract's conclusion is based on the 3- and 4-prong panels.
[5; Abstract] The abstract states without qualification that Planing and PCA 'deliver similar performance' to adversarial and uBoost. This is a statement about the specific 11-variable N-subjettiness representation of Eq. (4); Sec. 5 explicitly leaves other representations to future work. Because the relative difficulty of decorrelation could depend on the input representation, the conclusion should either be phrased as representation-dependent or the abstract and conclusion should carry the qualifier 'for the N-subjettiness inputs studied here.'

minor comments (6)

[3.3 heading] The heading 'Adverserial Neural Networks' contains a typo ('Adverserial' should be 'Adversarial').
[4.3, Fig. 15] The abstract's 'similar performance' glosses over the 2-prong ordering; at high background rejection the adversarial network clearly has the smallest Bhattacharyya distance, and the Planing curve has lower signal efficiency at fixed rejection. Suggest a more nuanced summary.
[Appendix B] The phrase 'Bhattacharya distance' should be 'Bhattacharyya distance'.
[Eq. (9)] The quantity used is a normalized variant of the Bhattacharyya distance; please cite a source or add a phrase such as 'normalized Bhattacharyya distance' to avoid confusion with the standard −ln BC definition.
[Figs. 12 and 14, bottom rows] The background-rejection axes are plotted with decreasing numerical values; consider reversing the axis or adding an arrow to clarify the 'better' direction.
[6, Conclusion] The statement 'Code to reproduce our results can be found on GitHub' would benefit from a URL or repository identifier.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports measured, head-to-head comparisons of existing decorrelation methods and does not derive its headline claim from its own inputs.

full rationale

The paper's central claim is an empirical benchmark: that data-augmentation methods (Planing and PCA rescaling) achieve comparable background-decorrelation performance to training-augmentation methods (adversarial networks and uBoost) with lower computational cost. This claim is supported by ROC curves, Bhattacharyya distances, and training-time measurements (e.g., Figs. 12-15 and Table 2), all obtained by actually training the classifiers on simulated QCD and signal jets and evaluating them on a common test set. No quantity asserted as an output is used as an input by construction. The planing weights in Eq. (6) are defined to flatten the training mass distribution, but the paper does not use that definition to 'predict' the post-training sculpting; the sculpting is measured afterward. The PCA transformation in Eq. (7) is applied as a preprocessing step, and again the resulting background rejection and Bhattacharyya distances are measured, not derived from the transformation itself. The only self-citation elements are Refs. [7] (PCA) and [9] (Planing), which are co-authored by two of the present authors. However, the paper implements those published algorithms as tools; the comparison does not rely on the conclusions of those prior papers as evidence. Thus the self-citations are not load-bearing. The lack of error bars on the adversarial training runs, noted in the Table 2 footnote ('statistics are over sampled once over each of the nine different values of lambda due to the long training time'), is a legitimate statistical robustness concern about the strength of the 'nearly identical' comparison, but it is not a circularity defect. The derivation chain is self-contained: inputs are simulated events, hyperparameters are stated, and all reported performance metrics come from independent testing after training.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new entities. Its free parameters are hyperparameters of the methods being compared, not fitted to produce the central claim. The assumptions are standard MC and ML practice.

free parameters (5)

Adversarial weight lambda = 50
Chosen by the authors from a scan over {1, 2, 5, 10, 20, 50, 100, 200, 500, 1000}; saturation at 50 is claimed (Sec. 3.3.2, App. A). The adversarial results in the comparison depend on this choice.
uBoost parameter beta_u = 1
Set by the authors following Ref. [30]; controls the relative importance of events in the boosted training (Sec. 3.3.1).
uBoost nearest-neighbor count k = 50
Chosen for local efficiency estimation; the paper notes little change for k in [50, 1000] (Sec. 3.3.1).
Number of BDTs in uBoost = 20
Covers the signal efficiency range; results are interpolated between target efficiencies (Sec. 3.3.1).
Learning rate and architecture sizes = 0.1 (BDT), 1e-3 (NN), 3 hidden layers x 50 nodes
Standard hyperparameters that saturate performance, chosen following cited references and author experiments (Sec. 3.1).

assumptions (4)

domain assumption The Monte Carlo event generation (MadGraph, Pythia, Delphes) accurately models the QCD background and signal jet substructure.
The entire comparison is based on simulated data; if the simulations mis-model mass-substructure correlations, the measured sculpting would be wrong. Invoked throughout Sec. 2.
domain assumption The 11-variable N-subjettiness basis (Eq. 4) spans the 5-body phase space and captures the information relevant for classifying up to 4-prong jets.
Used as the common input to all multivariate classifiers (Sec. 3.1). The paper cites Datta and Larkoski [44] for completeness.
domain assumption The alternating adversarial training procedure converges to a useful optimum.
The adversarial network training scheme is a heuristic procedure (Sec. 3.3.2); no proof of convergence is provided.
domain assumption Background rejection evaluated over the full mass window 50-400 GeV is representative of narrower signal windows.
Used for all ROC and distance metrics; the paper asserts this without a detailed demonstration (Sec. 4).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mass Agnostic Jet Taggers." pith.science (2026). https://pith.science/paper/DRTCBNHW

@misc{pith2026190808959,
  author       = {Pith},
  title        = {Pith review of: Mass Agnostic Jet Taggers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DRTCBNHW}},
  note         = {Machine review of arXiv:1908.08959}
}

read the original abstract

Searching for new physics in large data sets needs a balance between two competing effects---signal identification vs background distortion. In this work, we perform a systematic study of both single variable and multivariate jet tagging methods that aim for this balance. The methods preserve the shape of the background distribution by either augmenting the training procedure or the data itself. Multiple quantitative metrics to compare the methods are considered, for tagging 2-, 3-, or 4-prong jets from the QCD background. This is the first study to show that the data augmentation techniques of Planing and PCA based scaling deliver similar performance as the augmented training techniques of Adversarial NN and uBoost, but are both easier to implement and computationally cheaper.

Figures

Figures reproduced from arXiv: 1908.08959 by the authors.

**Figure 2.** The architecture of a BDT. We take the BDT to be made of 150 DTs, with a max depth of 4. The input to the BDT are the variables that span the 5-body jet phase space, see Eq. (4). The indicated parameters αi represent the weight associated with the particular DT. decision tree and neural network architectures, which have been shown to lead to increased discrimination.3 The authors of [44] introduced a minimal but com… view at source ↗

**Figure 3.** Many of the methods explored in this paper use a neural network classifier. For consistency, we always use a network with three hidden layers, each of which has 50 nodes and uses the ReLu activation function. The output is a single node with a sigmoid activation function. Our input data are the 11 τ (β) N variables of 5-body jet phase space shown in Eq. (4). A summary of the application of the three different method… view at source ↗

Figures from the paper (18 more)

**Figure 4.** Figure 4: The left panel shows the ROC curves for three traditional methods, two based on machine learning, to classify a 2-prong signal jet from a QCD jet. The machine learning based methods achieve an area significantly higher than the single variable τ21 based classifier. The…

**Figure 5.** Figure 5: The right panels of Fig. 5 show how the background distribution changes as tighter [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

**Figure 5.** Figure 5: The left panel shows the ROC curves for the data augmented neural network methods of PCA and planing as well as the single variable DDT. The network trained on PCA-rescaled data is the best classifier, followed by the network trained on planed data. Both MV decorrelati…

**Figure 6.** Figure 6: The upper and lower panels show distributions before and after planing away the jet mass, respectively. The left panels show the jet mass distribution for the 2-pronged signal and QCD background. By design, both distributions are (nearly) identical, and uniform across …

**Figure 7.** Figure 7: Scatterplot of two benchmark τ variables for QCD events in three different mass windows. The left panel shows the original variables, before any kind of preprocessing. The events from different mass bins are well separated. The center panel shows the same events after …

**Figure 8.** Figure 8: The network architecture used in the uBoost algorithm. Each BDT has the same layout as those in [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]

**Figure 9.** Figure 9: The left panel shows the ROC curves for the adversarially trained neural network and uBoost, along with the results of the base neural network and τ21, for comparison. The adversarial results use λ = 50, and the uBoost results use βu = 1. The right panels show the back…

**Figure 10.** Figure 10: The setup of our adversarially trained neural network. The classifier has the same hyperparameters as in [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

**Figure 11.** Figure 11: ROC curves for the 2-, 3-, and 4-prong signal jets versus QCD background for the methods which augment the training method to decorrelate the jet mass. The solid, dashed, and dotted curves show results for neural networks, boosted decision trees, and single variable a…

**Figure 12.** Figure 12: The Bhattacharyya distance for the QCD background distributions compared to the original distributions. The distance is defined in Eq. (9), and a larger distance represents more sculpting— lower on the plot is better. The upper and lower rows plot the distance as a f…

**Figure 13.** Figure 13: ROC curves for the 2-, 3-, and 4-prong signal jets versus QCD background for the methods which augment the data to decorrelate the jet mass rather than augment the training. The dashed and solid lines show the gradient boosted decision trees (BDT) and neural networks …

**Figure 14.** Figure 14: The Bhattacharyya distance for the QCD background distributions compared to the original distributions. The distance is defined in Eq. (9), and a larger distance represents more sculpting— lower on the plot is better. The neural networks tend to sculpt the distributi…

**Figure 15.** Figure 15 [PITH_FULL_IMAGE:figures/full_fig_p025_15.png]

**Figure 16.** Figure 16: The top row shows the ROC curves for all of the adversarially-trained neural networks tasked with distinguishing the 2-, 3-, and 4-prong signal jets from the QCD background. Lighter shades correspond to increasingly larger values of λ. Larger values of λ put an increa…

**Figure 17.** Figure 17: Comparison of Bhattacharyya distance and Jensen-Shannon distance for 2-, 3-, and 4- pronged signals, as a function of signal efficiency for various decorrelation methods studied in this work. The general trend for both metrics is seen to be the same. 30 [PITH_FULL_IM…

**Figure 18.** Figure 18: Comparison of all decorrelation methods to the benchmarks for the 2-prong signal. τN /τN−1 is τ2/τ1. 32 [PITH_FULL_IMAGE:figures/full_fig_p032_18.png]

**Figure 19.** Figure 19: Comparison of all decorrelation methods to the benchmarks for the 3-prong signal. τN /τN−1 is τ3/τ2. 33 [PITH_FULL_IMAGE:figures/full_fig_p033_19.png]

**Figure 20.** Figure 20: Comparison of all decorrelation methods to the benchmarks for the 4-prong signal. τN /τN−1 is τ4/τ3. 34 [PITH_FULL_IMAGE:figures/full_fig_p034_20.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Higgs Signal Strength Estimation with Machine Learning under Systematic Uncertainties
hep-ph 2025-08 conditional novelty 6.0 of 10

SAGE, a dual-branch GNN trained under nuisance fluctuations, estimates the Higgs signal strength with near-nominal coverage (0.662-0.683) but wider intervals than the top FAIR-HUC leaderboard methods.
Improving the performance of weak supervision searches using data augmentation
hep-ph 2024-11 conditional novelty 4.0 of 10

Physics-inspired data augmentation halves the signal data requirement for CWoLa weak supervision searches, cutting the practical sensitivity threshold from roughly 6 sigma to roughly 3 sigma.

Reference graph

Works this paper leans on

50 extracted references · 11 canonical work pages · cited by 2 Pith papers

[1]

Maximizing Boosted Top Identiﬁcation by Minimizing N-subjettiness,

J. Thaler and K. Van Tilburg, “Maximizing Boosted Top Identiﬁcation by Minimizing N-subjettiness,” JHEP 02 (2012) 093, arXiv:1108.2701 [hep-ph]

arXiv 2012
[2]

Identifying Boosted Objects with N-subjettiness,

J. Thaler and K. Van Tilburg, “Identifying Boosted Objects with N-subjettiness,” JHEP 03 (2011) 015, arXiv:1011.2268 [hep-ph]

arXiv 2011
[3]

Playing Tag with ANN: Boosted Top Identiﬁcation with Pattern Recognition,

L. G. Almeida, M. Backovi´ c, M. Cliche, S. J. Lee, and M. Perelstein, “Playing Tag with ANN: Boosted Top Identiﬁcation with Pattern Recognition,” JHEP 07 (2015) 086, arXiv:1501.05968 [hep-ph]

arXiv 2015
[4]

Jet-images — deep learning edition,

L. de Oliveira, M. Kagan, L. Mackey, B. Nachman, and A. Schwartzman, “Jet-images — deep learning edition,” JHEP 07 (2016) 069, arXiv:1511.05190 [hep-ph]

arXiv 2016
[5]

Parameterized neural networks for high-energy physics,

P. Baldi, K. Cranmer, T. Faucett, P. Sadowski, and D. Whiteson, “Parameterized neural networks for high-energy physics,” Eur. Phys. J. C76 (2016) no. 5, 235, arXiv:1601.07913 [hep-ex]

arXiv 2016
[6]

Identification of High-Momentum Top Quarks, Higgs Bosons, and W and Z Bosons Using Boosted Event Shapes

J. S. Conway, R. Bhaskar, R. D. Erbacher, and J. Pilot, “Identiﬁcation of High-Momentum Top Quarks, Higgs Bosons, and W and Z Bosons Using Boosted Event Shapes,” Phys. Rev. D94 (2016) no. 9, 094027, arXiv:1606.06859 [hep-ex]

work page Pith review arXiv 2016
[7]

A generic anti-QCD jet tagger,

J. A. Aguilar-Saavedra, J. H. Collins, and R. K. Mishra, “A generic anti-QCD jet tagger,” JHEP 11 (2017) 163, arXiv:1709.01087 [hep-ph]

arXiv 2017
[8]

Deep-learned Top Tagging with a Lorentz Layer,

A. Butter, G. Kasieczka, T. Plehn, and M. Russell, “Deep-learned Top Tagging with a Lorentz Layer,” SciPost Phys. 5 (2018) no. 3, 028, arXiv:1707.08966 [hep-ph]

arXiv 2018

Show all 50 references

[9]

What is the Machine Learning?,

S. Chang, T. Cohen, and B. Ostdiek, “What is the Machine Learning?,” Phys. Rev. D97 (2018) no. 5, 056009, arXiv:1709.10106 [hep-ph]

2018 arXiv
[10]

(Machine) Learning to Do More with Less,

T. Cohen, M. Freytsis, and B. Ostdiek, “(Machine) Learning to Do More with Less,” JHEP 02 (2018) 034, arXiv:1706.09451 [hep-ph]

2018 arXiv
[11]

Weakly Supervised Classiﬁcation in High Energy Physics,

L. M. Dery, B. Nachman, F. Rubbo, and A. Schwartzman, “Weakly Supervised Classiﬁcation in High Energy Physics,” JHEP 05 (2017) 145, arXiv:1702.00414 [hep-ph]

2017 arXiv
[12]

Long Short-Term Memory (LSTM) networks with jet constituents for boosted top tagging at the LHC,

S. Egan, W. Fedorko, A. Lister, J. Pearkes, and C. Gay, “Long Short-Term Memory (LSTM) networks with jet constituents for boosted top tagging at the LHC,” arXiv:1711.09059 [hep-ex]

arXiv
[13]

Deep-learning Top Taggers or The End of QCD?,

G. Kasieczka, T. Plehn, M. Russell, and T. Schell, “Deep-learning Top Taggers or The End of QCD?,” JHEP 05 (2017) 006, arXiv:1701.08784 [hep-ph]

2017 arXiv
[14]

Infrared Safety of a Neural-Net Top Tagging Algorithm,

S. Choi, S. J. Lee, and M. Perelstein, “Infrared Safety of a Neural-Net Top Tagging Algorithm,” JHEP 02 (2019) 132, arXiv:1806.01263 [hep-ph]

2019 arXiv
[15]

Anomaly Detection for Resonant New Physics with Machine Learning,

J. H. Collins, K. Howe, and B. Nachman, “Anomaly Detection for Resonant New Physics with Machine Learning,” Phys. Rev. Lett. 121 (2018) no. 24, 241803, arXiv:1805.02664 [hep-ph] . 35 SciPost Physics Submission

2018 arXiv
[16]

Jet Charge and Machine Learning,

K. Fraser and M. D. Schwartz, “Jet Charge and Machine Learning,” JHEP 10 (2018) 093, arXiv:1803.08066 [hep-ph]

2018 arXiv
[17]

Learning to classify from impure samples with high-dimensional data,

P. T. Komiske, E. M. Metodiev, B. Nachman, and M. D. Schwartz, “Learning to classify from impure samples with high-dimensional data,” Phys. Rev. D98 (2018) no. 1, 011502, arXiv:1801.10158 [hep-ph]

2018 arXiv
[18]

Spectral Analysis of Jet Substructure with Neural Networks: Boosted Higgs Case,

S. H. Lim and M. M. Nojiri, “Spectral Analysis of Jet Substructure with Neural Networks: Boosted Higgs Case,” JHEP 10 (2018) 181, arXiv:1807.03312 [hep-ph]

2018 arXiv
[19]

Boosting H→b¯b with Machine Learning,

J. Lin, M. Freytsis, I. Moult, and B. Nachman, “Boosting H→b¯b with Machine Learning,” JHEP 10 (2018) 101, arXiv:1807.10768 [hep-ph]

2018 arXiv
[20]

Pulling Out All the Tops with Computer Vision and Deep Learning,

S. Macaluso and D. Shih, “Pulling Out All the Tops with Computer Vision and Deep Learning,” JHEP 10 (2018) 121, arXiv:1803.00107 [hep-ph]

2018 arXiv
[21]

Searching for New Physics with Deep Autoencoders,

M. Farina, Y. Nakai, and D. Shih, “Searching for New Physics with Deep Autoencoders,” arXiv:1808.08992 [hep-ph]

arXiv
[22]

Extending the search for new resonances with machine learning,

J. H. Collins, K. Howe, and B. Nachman, “Extending the search for new resonances with machine learning,” Phys. Rev. D99 (2019) no. 1, 014038, arXiv:1902.02634 [hep-ph]

2019 arXiv
[23]

Automating the Construction of Jet Observables with Machine Learning,

K. Datta, A. Larkoski, and B. Nachman, “Automating the Construction of Jet Observables with Machine Learning,” arXiv:1902.07180 [hep-ph]

1902 arXiv
[24]

The Machine Learning Landscape of Top Taggers,

A. Butter et al. , “The Machine Learning Landscape of Top Taggers,” arXiv:1902.09914 [hep-ph]

1902 arXiv
[25]

Jet Substructure at the Large Hadron Collider: A Review of Recent Advances in Theory and Machine Learning,

A. J. Larkoski, I. Moult, and B. Nachman, “Jet Substructure at the Large Hadron Collider: A Review of Recent Advances in Theory and Machine Learning,” arXiv:1709.04464 [hep-ph]

arXiv
[26]

Deep Learning and its Application to LHC Physics,

D. Guest, K. Cranmer, and D. Whiteson, “Deep Learning and its Application to LHC Physics,” Ann. Rev. Nucl. Part. Sci. 68 (2018) 161–181, arXiv:1806.11484 [hep-ex]

2018 arXiv
[27]

Decorrelated Jet Substructure Tagging using Adversarial Neural Networks,

C. Shimmin, P. Sadowski, P. Baldi, E. Weik, D. Whiteson, E. Goul, and A. Søgaard, “Decorrelated Jet Substructure Tagging using Adversarial Neural Networks,” Phys. Rev. D96 (2017) no. 7, 074034, arXiv:1703.03507 [hep-ex]

2017 arXiv
[28]

Thinking outside the ROCs: Designing Decorrelated Taggers (DDT) for jet substructure,

J. Dolen, P. Harris, S. Marzani, S. Rappoccio, and N. Tran, “Thinking outside the ROCs: Designing Decorrelated Taggers (DDT) for jet substructure,” JHEP 05 (2016) 156, arXiv:1603.00027 [hep-ph]

2016 arXiv
[29]

Convolved Substructure: Analytically Decorrelating Jet Substructure Observables,

I. Moult, B. Nachman, and D. Neill, “Convolved Substructure: Analytically Decorrelating Jet Substructure Observables,” JHEP 05 (2018) 002, arXiv:1710.06859 [hep-ph]

2018 arXiv
[30]

uBoost: A boosting method for producing uniform selection eﬃciencies from multivariate classiﬁers,

J. Stevens and M. Williams, “uBoost: A boosting method for producing uniform selection eﬃciencies from multivariate classiﬁers,” JINST 8 (2013) P12013, arXiv:1305.7248 [nucl-ex] . 36 SciPost Physics Submission

2013 arXiv
[31]

Learning to Pivot with Adversarial Networks,

G. Louppe, M. Kagan, and K. Cranmer, “Learning to Pivot with Adversarial Networks,” arXiv:1611.01046 [stat.ME]

arXiv
[32]

QCD or What?,

T. Heimel, G. Kasieczka, T. Plehn, and J. M. Thompson, “QCD or What?,” SciPost Phys. 6 (2019) no. 3, 030, arXiv:1808.08979 [hep-ph]

2019 arXiv
[33]

Performance of mass-decorrelated jet substructure observables for hadronic two-body decay tagging in ATLAS,

ATLAS Collaboration Collaboration, “Performance of mass-decorrelated jet substructure observables for hadronic two-body decay tagging in ATLAS,” Tech. Rep. ATL-PHYS-PUB-2018-014, CERN, Geneva, Jul, 2018. https://cds.cern.ch/record/2630973

2018
[34]

Flavor Universal Resonances and Warped Gravity,

K. Agashe, P. Du, S. Hong, and R. Sundrum, “Flavor Universal Resonances and Warped Gravity,” JHEP 01 (2017) 016, arXiv:1608.00526 [hep-ph]

2017 arXiv
[35]

LHC Signals from Cascade Decays of Warped Vector Resonances,

K. S. Agashe, J. Collins, P. Du, S. Hong, D. Kim, and R. K. Mishra, “LHC Signals from Cascade Decays of Warped Vector Resonances,” JHEP 05 (2017) 078, arXiv:1612.00047 [hep-ph]

2017 arXiv
[36]

Dedicated Strategies for Triboson Signals from Cascade Decays of Vector Resonances,

K. Agashe, J. H. Collins, P. Du, S. Hong, D. Kim, and R. K. Mishra, “Dedicated Strategies for Triboson Signals from Cascade Decays of Vector Resonances,” Phys. Rev. D99 (2019) no. 7, 075016, arXiv:1711.09920 [hep-ph]

2019 arXiv
[37]

Detecting a Boosted Diboson Resonance,

K. Agashe, J. H. Collins, P. Du, S. Hong, D. Kim, and R. K. Mishra, “Detecting a Boosted Diboson Resonance,” JHEP 11 (2018) 027, arXiv:1809.07334 [hep-ph]

2018 arXiv
[38]

The automated computation of tree-level and next-to-leading order diﬀerential cross sections, and their matching to parton shower simulations,

J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, O. Mattelaer, H. S. Shao, T. Stelzer, P. Torrielli, and M. Zaro, “The automated computation of tree-level and next-to-leading order diﬀerential cross sections, and their matching to parton shower simulations,” JHEP 0...

2014 arXiv
[39]

A Brief Introduction to PYTHIA 8.1,

T. Sjostrand, S. Mrenna, and P. Z. Skands, “A Brief Introduction to PYTHIA 8.1,” Comput. Phys. Commun. 178 (2008) 852–867, arXiv:0710.3820 [hep-ph]

2008 arXiv
[40]

DELPHES 3, A modular framework for fast simulation of a generic collider experiment,

DELPHES 3 Collaboration, J. de Favereau, C. Delaere, P. Demin, A. Giammanco, V. Lemaˆ ıtre, A. Mertens, and M. Selvaggi, “DELPHES 3, A modular framework for fast simulation of a generic collider experiment,” JHEP 02 (2014) 057, arXiv:1307.6346 [hep-ex]

2014 arXiv
[41]

Techniques for improved heavy particle searches with jet substructure,

S. D. Ellis, C. K. Vermilion, and J. R. Walsh, “Techniques for improved heavy particle searches with jet substructure,” Phys. Rev. D80 (2009) 051501, arXiv:0903.5081 [hep-ph]

2009 arXiv
[42]

N-Jettiness: An Inclusive Event Shape to Veto Jets,

I. W. Stewart, F. J. Tackmann, and W. J. Waalewijn, “N-Jettiness: An Inclusive Event Shape to Veto Jets,” Phys. Rev. Lett. 105 (2010) 092002, arXiv:1004.2489 [hep-ph]

2010 arXiv
[43]

Stealth multiboson signals,

J. A. Aguilar-Saavedra, “Stealth multiboson signals,” Eur. Phys. J. C77 (2017) no. 10, 703, arXiv:1705.07885 [hep-ph]

2017 arXiv
[44]

How Much Information is in a Jet?,

K. Datta and A. Larkoski, “How Much Information is in a Jet?,” JHEP 06 (2017) 073, arXiv:1704.08249 [hep-ph] . 37 SciPost Physics Submission

2017 arXiv
[45]

Scikit-learn: Machine learning in Python,

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, “Scikit-learn: Machine learning in Python,” Journal of Machine Learni...

2011
[46]

Chollet et al

F. Chollet et al. , “Keras.” https://keras.io, 2015

2015
[47]

TensorFlow: Large-scale machine learning on heterogeneous systems,

M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Isard, Y. Jia, R. Jozefowicz, L. Kaiser, M. Kudlur, J. Levenberg, D. Man´ e, R. Monga, S. Moore, D. Murray, C. Olah, M....

2015
[48]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv:1412.6980 (2014)

2014 arXiv
[49]

Neural Networks for Full Phase-space Reweighting and Parameter Tuning,

A. Andreassen and B. Nachman, “Neural Networks for Full Phase-space Reweighting and Parameter Tuning,” arXiv:1907.08209 [hep-ph]

1907 arXiv
[50]

New approaches for boosting to uniformity,

A. Rogozhnikov, A. Bukva, V. V. Gligorov, A. Ustyuzhanin, and M. Williams, “New approaches for boosting to uniformity,” JINST 10 (2015) no. 03, T03002, arXiv:1410.4140 [hep-ex] . 38

2015 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Maximizing Boosted Top Identiﬁcation by Minimizing N-subjettiness,

J. Thaler and K. Van Tilburg, “Maximizing Boosted Top Identiﬁcation by Minimizing N-subjettiness,” JHEP 02 (2012) 093, arXiv:1108.2701 [hep-ph]

arXiv 2012

[2] [2]

Identifying Boosted Objects with N-subjettiness,

J. Thaler and K. Van Tilburg, “Identifying Boosted Objects with N-subjettiness,” JHEP 03 (2011) 015, arXiv:1011.2268 [hep-ph]

arXiv 2011

[3] [3]

Playing Tag with ANN: Boosted Top Identiﬁcation with Pattern Recognition,

L. G. Almeida, M. Backovi´ c, M. Cliche, S. J. Lee, and M. Perelstein, “Playing Tag with ANN: Boosted Top Identiﬁcation with Pattern Recognition,” JHEP 07 (2015) 086, arXiv:1501.05968 [hep-ph]

arXiv 2015

[4] [4]

Jet-images — deep learning edition,

L. de Oliveira, M. Kagan, L. Mackey, B. Nachman, and A. Schwartzman, “Jet-images — deep learning edition,” JHEP 07 (2016) 069, arXiv:1511.05190 [hep-ph]

arXiv 2016

[5] [5]

Parameterized neural networks for high-energy physics,

P. Baldi, K. Cranmer, T. Faucett, P. Sadowski, and D. Whiteson, “Parameterized neural networks for high-energy physics,” Eur. Phys. J. C76 (2016) no. 5, 235, arXiv:1601.07913 [hep-ex]

arXiv 2016

[6] [6]

Identification of High-Momentum Top Quarks, Higgs Bosons, and W and Z Bosons Using Boosted Event Shapes

J. S. Conway, R. Bhaskar, R. D. Erbacher, and J. Pilot, “Identiﬁcation of High-Momentum Top Quarks, Higgs Bosons, and W and Z Bosons Using Boosted Event Shapes,” Phys. Rev. D94 (2016) no. 9, 094027, arXiv:1606.06859 [hep-ex]

work page Pith review arXiv 2016

[7] [7]

A generic anti-QCD jet tagger,

J. A. Aguilar-Saavedra, J. H. Collins, and R. K. Mishra, “A generic anti-QCD jet tagger,” JHEP 11 (2017) 163, arXiv:1709.01087 [hep-ph]

arXiv 2017

[8] [8]

Deep-learned Top Tagging with a Lorentz Layer,

A. Butter, G. Kasieczka, T. Plehn, and M. Russell, “Deep-learned Top Tagging with a Lorentz Layer,” SciPost Phys. 5 (2018) no. 3, 028, arXiv:1707.08966 [hep-ph]

arXiv 2018