Pith. sign in

REVIEW 4 major objections 6 minor 61 references

Unsupervised Domain Adaptation for Multitask Image Analysis in Realistic Context with Extreme Label Shift; Application to the CTAO first Large Sized Telescope

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper argues that conditional domain-adversarial training combined with uncertainty weighting is the most promising unsupervised domain adaptation plus multitask configuration for CTAO LST-1 gamma-ray image analysis under extreme…

desk verdict Useful, code-sharing engineering study of conditional UDA under extreme label shift, but the CTAO conclusion that CDANN+UW wins is not statistically supported by single-run, target-test-selected numbers. read the letter →

arxiv 2608.09630 v1 pith:EDDCCCOF submitted 2026-08-10 astro-ph.IM cs.CV

Michaël Dell'aiera , Thomas Vuillaume , Alexandre Benoit This is my paper · ORCID
classification astro-ph.IMcs.CV
keywords unsuperviseddomainadaptationmultitasklearninglabelshiftimportanceweightingconditionalgamma-rayastronomyCherenkovTelescopeArraynightskybackground
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

The paper proposes combining unsupervised domain adaptation (UDA) with automatic multitask balancing so that a deep network trained on labeled gamma-ray simulations can be adapted to unlabeled real telescope images despite an extreme class imbalance. Its central claim, stated in the conclusion, is that the most promising configuration for the CTAO first Large-Sized Telescope is conditional domain-adversarial training (CDANN) paired with uncertainty weighting (UW). The paper builds a controlled target dataset by adding Poisson noise at rate 0.46 to simulated images and fixing a gamma/proton ratio of $10^{-4}$, then compares DANN, DeepCORAL and DeepJDOT, each with and without conditioning and with manual, uncertainty or GradNorm balancing. On digit benchmarks it shows automatic balancing matches or beats manual weight tuning, and on the LST simulations the conditional variants recover performance lost to label shift.

What carries the argument

The central object is the conditional domain-adaptation loss, a modification of three UDA objectives (DANN, DeepJDOT, DeepCORAL) that restricts the domain-alignment term to source protons paired with unlabeled target samples, so the extreme gamma/proton label shift does not poison the adaptation. A second mechanism is the gradient layer (GL), a scheduler that weights the domain-adaptation gradient during backpropagation and was found helpful for DeepJDOT and DeepCORAL but harmful for DANN. The third is automatic multitask balancing: uncertainty weighting (UW), which sets loss weights from task-dependent homoscedastic uncertainty, and GradNorm (GN), which sets them from gradient norms and inverse training rates.

What would settle it

Apply the same benchmark to real LST-1 observations of the Crab Nebula, using the Monte-Carlo simulations as labeled source and the real data as unlabeled target, then compare CDANN+UW against the non-conditional and GradNorm variants with the paper's IRF metrics; if CDANN+UW does not improve AUC, energy bias, or angular resolution over the alternatives, the simulation-based ranking is falsified. A cheaper check is to compare the distribution of cleaned image moments between real data and Poisson-degraded simulations; a mismatch beyond the 0.46 noise level would break the controlled-target premise that supports the ranking.

Watch

Extended reading notes

Core claim

The paper's core claim is that the dominant discrepancy between simulations and telescope observations can be handled by a conditional UDA loss that aligns only source protons with the unlabeled target batch, using the known gamma/proton ratio as an importance weight. For DANN, conditioning means the domain-classifier loss is masked so that source gammas do not participate in the alignment; for DeepJDOT and DeepCORAL the analogous step restricts the transport cost and the covariance statistics to source protons. The paper reports that CDANN combined with uncertainty weighting yields the best energy bias and resolution on the controlled target, and that the conditional variants of DeepCORAL and DeepJDOT improve energy reconstruction only when paired with GradNorm. It also reports that DeepJDOT, while theoretically attractive, is unstable and impractical at CTAO scale because of quadratic cost and large initial gradients. The authors explicitly restrict the claim to controlled simulations and say results on real telescope data will be presented in a subsequent work.

Load-bearing premise

The paper's ranking of methods rests on the assumption that adding Poisson noise at rate 0.46 and fixing the gamma/proton ratio at $10^{-4}$ reproduces the main gap between simulated and real telescope images; if real observations differ in other ways, the winning method could change.

Editorial extensions

If this is right

  • CDANN+UW, if correct, becomes the recommended configuration for adapting CTAO LST-1 models trained on simulations to real unlabeled observations under night-sky background and extreme label shift.
  • Conditional UDA makes the adaptation stable even when the target gamma/proton ratio is as low as 10^-4, so the method does not require knowing the exact source flux in advance.
  • Automatic balancing with UW or GN removes the grid search over task weights and, on digit benchmarks, matches or outperforms manual weighting.
  • DeepJDOT's quadratic optimal-transport cost and training instability make it less suited than CDANN for large-scale astrophysical datasets.
  • The ranking is currently validated only on Poisson-degraded simulations at delta-lambda=0.46 and ratio 10^-4, so real-data validation is the explicit next step.

Reading between the lines

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

  • The paper does not test whether aligning only source protons is enough when real gamma events have different morphology than simulated ones; a gamma-specific residual shift would remain uncorrected if that difference is large.
  • Because the AUC stays nearly constant across target ratios from 10^-1 to 10^-5 in the paper's Table 6, the method may tolerate fairly large errors in the estimated gamma/proton ratio, which is a testable property for real observations.
  • A natural extension, which the paper does not run, is to add a second degradation such as a modified point-spread function or atmospheric absorption to see whether CDANN+UW remains the best configuration when the shift is not purely Poisson.
  • The gradient layer only scales gradient amplitude and does not resolve direction conflicts; an editorial next step is to combine conditional UDA with gradient-surgery balancing, which the paper mentions as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes to combine unsupervised domain adaptation (UDA) with multitask balancing (MTB), introducing conditional versions of DANN, DeepJDOT, and DeepCORAL that restrict the adaptation objective to the majority source class in order to handle extreme label shift. The methods are first validated on digit benchmarks (MNIST, USPS, SVHN, MNISTM) with ten-seed averages, and then applied to a controlled CTAO LST-1 simulation setup in which the target domain is built by adding Poisson noise (δλ = 0.46) to simulated images and imposing a gamma/proton ratio of 10^-4. The central claim, stated in the conclusion, is that CDANN paired with Uncertainty Weighting (UW) is the most promising configuration for real gamma-ray source detection and analysis. The paper also introduces a Gradient Layer (GL) scheduler for domain-adaptation gradients and publishes code and results on Zenodo.

Significance. If the central claim holds, the paper provides a practical recipe for applying UDA to gamma-ray astronomy under extreme label shift, a regime that is rarely studied and is highly relevant for CTAO. The controlled simulation framework with ground truth labels is well chosen, and the comparison covers three UDA families (adversarial, optimal-transport, and correlation alignment) with multiple MTB strategies. The digit experiments use ten seeds, and the paper explicitly reports convergence issues and hyperparameter sensitivities, which is honest. The open-source release of code and results is a concrete strength. The significance is currently limited by the lack of statistical grounding in the CTAO experiments and by the use of the target test set for model selection, both of which weaken the support for the headline ranking.

major comments (4)
  1. [Section 5.4.4, Tables 7–9] The CTAO ablation results are reported as single numbers without error bars or run counts, whereas the digit experiments are averaged over ten seeds (Section 4.2). The differences that motivate the selection of CDANN+UW over CDANN+GN (e.g., Table 7: E_mu 0.08 vs 0.09, E_sigma 0.31 vs 0.33, theta_sigma 0.26 vs 0.27, AUC 0.82 vs 0.82) are within the typical seed-to-seed variability of deep multitask networks. The authors should repeat each CTAO configuration with multiple seeds and report means with standard deviations or confidence intervals. Without this, the ranking that drives the paper's central claim is not statistically supported.
  2. [Section 5.4.3 and Tables 7–9] The GradNorm hyperparameter alpha and the final winning method are selected using metrics computed on the target test set, which is the same set used to report the headline numbers. This is selection-on-test: the comparison is biased in favor of the chosen configuration, and the reported edge over alternatives is optimistic. To support the conclusion, the authors should either hold out a target validation set for all hyperparameter decisions and report results on a separate target test set, or explicitly state that the reported figures are the result of oracle (test-set) selection and should be interpreted with that caveat.
  3. [Section 4.3, Table 2] The DeepJDOT baseline does not reproduce the performance reported in the original DeepJDOT paper [25]; on SVHN→MNIST, the authors report accuracies of 0.65–0.75 depending on the MTB strategy, and state that they 'did not manage to retrieve the results of [25] in this particular scenario.' Since DeepJDOT is one of the three UDA baselines used in the CTAO comparison (Table 9 and Section 5.4.5), a baseline that is not validated against the published method makes the comparative conclusions potentially unfair. The authors should either tune DeepJDOT to reproduce the published digit results, or provide a detailed analysis of why their implementation is the correct one for a fair comparison in the CTAO setting.
  4. [Section 6 (Conclusion) vs Sections 5.4.1–5.4.5] The conclusion states that the comparative study was conducted on simulations with 'multiple NSB and label shift,' but the actual UDA method comparison (Tables 7–9) uses a single target configuration: δλ = 0.46 and a gamma/proton ratio of 10^-4. Table 5 varies δλ only for the non-adapted baseline model, and Table 6 varies the ratio only for DANN without the conditional or MTB variants. The ranking that selects CDANN+UW is therefore established at only one operating point. To support the conclusion, the authors should either present UDA results across the δλ range of Table 5 and across the ratio range of Table 6, or soften the claim to reflect the single-point comparison.
minor comments (6)
  1. [Eq. (12)–(13)] The definition of the gradient layer derivative is ambiguous: dG/dx = (−1)R f(i) × I should be written more clearly, e.g., dG/dx = −R·f(i)·I for DANN and dG/dx = +f(i)·I for DeepJDOT/DeepCORAL, or with the exponent written as (−1)^R · f(i). The current notation makes it hard to see which sign is used in each case.
  2. [Table 4] The 'performance gap' values in Table 4 are not tied to Table 2. For example, the +21.8 gap for DeepJDOT with UW is much larger than any difference visible in Table 2; the reader cannot tell whether Table 2 already includes the Gradient Layer or whether the gap is computed relative to a different base configuration. Please clarify the relationship between the two tables.
  3. [Section 2.4] The claim that 'there is no work trying to associate UDA and MTB using GN or UW' is a strong novelty statement that is not substantiated by a systematic literature search. Given the popularity of both techniques, it would be safer to soften the claim or provide a more thorough review of prior attempts at automatic loss balancing in UDA.
  4. [Section 5.4.3, Tables 7–9] The sentence 'all our methods have a better performance when paired with GN' is too broad: in Table 7, DANN with UW has AUC 0.83 versus 0.82 with GN, and after conditioning CDANN+UW outperforms CDANN+GN on E_mu and theta_sigma. The text later acknowledges the change, but the earlier sentence should be qualified to refer to the non-conditional setting and to the specific metrics.
  5. [Section 5.4.3] Minor grammatical issue: 'DeepCORAL and DeepJDOT fails at converging' should be 'fail to converge'.
  6. [Section 6] The phrase 'most promising tool for the detection and analysis of real gamma-ray sources' in the conclusion goes beyond the evidence, which is based on a controlled simulation target. The authors do add the caveat that real data will be studied later, but the wording of the central claim could be softened to 'most promising in the controlled simulation framework tested' to avoid overstating the external validity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the CDANN+UW selection is an empirical comparison on a controlled synthetic target; the self-citations provide architecture and setup context, not a fitted or definitional shortcut.

full rationale

The paper's derivation chain is empirical rather than tautological. The controlled target is created by adding Poisson noise (delta_lambda=0.46) and fixing the gamma/proton ratio at 1e-4; this is an experimental benchmark with ground-truth labels, not a quantity derived from the conclusion. The conditional methods (CDANN, CDeepJDOT, CDeepCORAL) are explicitly built on cited importance-weighting work [12,13] and are compared against their non-conditional counterparts and multiple baselines, so the reported gains are not forced by construction. The choice delta_lambda=0.46 is imported from previous work [6,51] as a setup parameter, not fitted to the target test metrics; the GN alpha sweep is a hyperparameter selection, and the final ablation tables are reported on the same target set, which is a test-set-selection/statistical-power concern but not a circularity. The paper's own caveat that 'additional discrepancies may be present in real observations' explicitly limits external validity and therefore does not hide circular reasoning. Self-citations [6,51] supply the gamma-PhysNet architecture and the prior UDA baseline, but the central claim (CDANN+UW is most promising in this controlled comparison) rests on the new ablation experiments, not on those citations being true by fiat. No equation in the paper is equivalent to its conclusion, and no fitted parameter is renamed as a prediction.

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

The central claim depends on a handful of hyperparameter choices and on domain assumptions about covariate shift, the Poisson-noise target model, and the extreme gamma/proton ratio. No new physical entities or exotic theoretical constructs are introduced; the main burden lies in the realism of the simulation-based target proxy and the test-set-driven selection of hyperparameters.

free parameters (4)
  • GradNorm exponent alpha = 0.1 for CTAO experiments
    Selected on the target evaluation set after grid search over {0.1, 0.5, 1.5, 3.0}; the paper reports better performance with smaller alpha, so this is a tuning choice rather than an a priori value.
  • Gradient layer steepness gamma = 10
    Set by the authors for the GL schedule; no ablation over gamma is shown, so it is a hand-chosen constant.
  • Target Poisson noise rate delta_lambda = 0.46
    Chosen according to previous work [51] to mimic NSB in CTAO observations; it sets the covariate shift magnitude and therefore the difficulty of the adaptation task.
  • Target gamma/proton ratio = 1e-4
    Assumed realistic value from [11]; the paper varies ratio in 5.4.2 but uses 1e-4 in the final comparisons, shaping the extreme label shift condition.
assumptions (4)
  • domain assumption Covariate shift assumption: P_S(y|x) = P_T(y|x)
    The paper relies on this invariance for UDA methods to be valid (Section 2.2). If the conditional label distribution changes between simulations and observations, the proposed alignment could be harmful.
  • ad hoc to paper The target domain is approximated by adding Poisson noise to simulated images with fixed rate delta_lambda=0.46
    Section 5.1 uses this degradation to create a proxy for real observations; the paper acknowledges it does not reproduce all discrepancies real data.
  • domain assumption The real gamma/proton ratio is at most 1e-4
    Section 3.1.3 states real data contain less than 1 gamma for 10^4 protons, drawing on [11]; this justifies the conditioning that removes gammas from the source alignment.
  • domain assumption Simulated events from CORSIKA/sim_telarray provide sufficiently accurate ground truth for evaluating reconstruction
    Section 5.1 relies on standard Monte Carlo simulations; the paper uses the same simulations for source and target (with noise added), so the evaluation does not test against independent real observations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unsupervised Domain Adaptation for Multitask Image Analysis in Realistic Context with Extreme Label Shift; Application to the CTAO first Large Sized Telescope." pith.science (2026). https://pith.science/paper/EDDCCCOF

@misc{pith2026260809630,
  author       = {Pith},
  title        = {Pith review of: Unsupervised Domain Adaptation for Multitask Image Analysis in Realistic Context with Extreme Label Shift; Application to the CTAO first Large Sized Telescope},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EDDCCCOF}},
  note         = {Machine review of arXiv:2608.09630}
}
read the original abstract

Unsupervised domain adaptation is a widespread set of methods that leverages the knowledge of a labeled source domain to train a model to perform well on a related unlabeled target domain. They generally introduce an auxiliary adaptation-related task that can be integrated into the multitask paradigm, which aims to merge multiple single-task models into a unified architecture. In this paper, we propose to associate domain adaptation and multitask balancing in the realistic context of an extreme class imbalance. Therefore, we propose a combined framework to cover and validate these approaches, and evaluate its performance in the physics-based context of the Cherenkov Telescope Array Observatory (CTAO). Along with a comparative study of some relevant adaptation techniques, we highlight the impact of extreme label shift and extend the investigations on importance weighting to rectify it. The complete code and results are published and available as open-source resources on Zenodo.

Figures

Figures reproduced from arXiv: 2608.09630 by the authors.

Figure 1
Figure 1. The γ-PhysNet architecture. This illustration represents the feature maps that output from the convolution filters or the fully-connected layers (that can be associated with a non-linearity) of the network. For example, conv designates the feature maps that output a convolution layer. The backbone is composed of three blocks and the identity mapping corresponds to full-activation [46]. in the presence of strong NSB,… view at source ↗
Figure 2
Figure 2. The cosine similarity between the domain task and the full-event reconstruction [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Illustration of UDA. While maintaining good performance on the source data [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Cosine similarity between the domain confusion task and the digit classification [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 20 canonical work pages

  1. [25]

    B. B. Damodaran, B. Kellenberger, R. Flamary, D. Tuia, N. Courty, Deepjdot: Deep joint distribution optimal transport for unsupervised domain adaptation (2018). doi:10.48550/ARXIV.1803.10081. URLhttps://arxiv.org/abs/1803.10081

    arXiv doi:10.48550/arxiv.1803.10081 2018

  2. [1]

    Dell’aiera, T

    M. Dell’aiera, T. Vuillaume, A. Benoit, 2024 domain adaptation meth- ods ctao lst-1 results (Sep. 2024). doi:10.5281/zenodo.13646001. URLhttps://doi.org/10.5281/zenodo.13646001 31

    work page doi:10.5281/zenodo.13646001 2024

  3. [2]

    LeCun, Y

    Y. LeCun, Y. Bengio, Convolutional Networks for Images, Speech, and Time Series, MIT Press, Cambridge, MA, USA, 1998, p. 255–258

    work page 1998

  4. [3]

    Torralba, A

    A. Torralba, A. A. Efros, Unbiased look at dataset bias, in: CVPR 2011, 2011, pp. 1521–1528. doi:10.1109/CVPR.2011.5995347

    arXiv 2011

  5. [4]

    Agostinelli, J

    S. Agostinelli, J. Allison, K. A. et al, Geant4—a simulation toolkit, Nuclear Instruments and Methods in Physics Research Section A: Ac- celerators, Spectrometers, Detectors and Associated Equipment 506 (3) (2003) 250–303. doi:https://doi.org/10.1016/S0168-9002(03)01368-8. URLhttps://www.sciencedirect.com/science/article/pii/ S0168900203013688

    doi:10.1016/s0168-9002(03)01368-8 2003

  6. [5]

    Bernlöhr, Simulation of imaging atmospheric cherenkov telescopes with corsika and sim_telarray, Astroparticle Physics 30 (3) (2008) 149–158

    K. Bernlöhr, Simulation of imaging atmospheric cherenkov telescopes with corsika and sim_telarray, Astroparticle Physics 30 (3) (2008) 149–158. doi:https://doi.org/10.1016/j.astropartphys.2008.07.009. URLhttps://www.sciencedirect.com/science/article/pii/ S0927650508000972

    doi:10.1016/j.astropartphys.2008.07.009 2008

  7. [6]

    Jacquemont, T

    M. Jacquemont, T. Vuillaume, A. Benoit, G. Maurin, P. Lambert, G. Lamanna, First full-event reconstruction from imaging atmospheric cherenkov telescope real data with deep learning, in: 2021 International Conference on Content-Based Multimedia Indexing (CBMI), 2021, pp. 1–6. doi:10.1109/CBMI50038.2021.9461918. URLhttps://ieeexplore.ieee.org/document/9461918

    arXiv 2021

  8. [7]

    S. Zhao, X. Yue, S. Zhang, B. Li, H. Zhao, B. Wu, R. Krishna, J. E. Gon- zalez, A. L. Sangiovanni-Vincentelli, S. A. Seshia, K. Keutzer, A review of single-source deep unsupervised visual domain adaptation (2020). arXiv:2009.00155

    work page Pith review arXiv 2020

Show all 61 references
  1. [8]

    Kendall, Y

    A. Kendall, Y. Gal, R. Cipolla, Multi-task learning using uncertainty to weigh losses for scene geometry and semantics (2018). arXiv:1705.07115

    2018 arXiv

  2. [9]

    Ganin, E

    Y. Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Lavio- lette, M. Marchand, V. Lempitsky, Domain-adversarial training of neu- ral networks (2016). arXiv:1505.07818

    2016 arXiv

  3. [10]

    R. D. Parsons, A. M. W. Mitchell, S. Ohm, Investigations of the sys- tematic uncertainties in convolutional neural network based analysis of atmospheric cherenkov telescope data (2022). arXiv:2203.05315. 32

    2022 arXiv

  4. [11]

    Abe, et al., Observations of the Crab Nebula and Pulsar with the Large-sized Telescope Prototype of the Cherenkov Telescope Array, Astrophys

    H. Abe, et al., Observations of the Crab Nebula and Pulsar with the Large-sized Telescope Prototype of the Cherenkov Telescope Array, Astrophys. J. 956 (2) (2023) 80. arXiv:2306.12960, doi:10.3847/1538- 4357/ace89d

    2023

  5. [12]

    X. Liu, Z. Guo, S. Li, F. Xing, J. You, C. C. J. Kuo, G. E. Fakhri, J. Woo, Adversarial unsupervised domain adaptation with conditional and label shift: Infer, align and iterate (2021). arXiv:2107.13469

    2021 arXiv

  6. [13]

    Z. C. Lipton, Y.-X. Wang, A. Smola, Detecting and correcting for label shift with black box predictors (2018). arXiv:1802.03916

    2018 arXiv

  7. [14]

    M. Kull, P. A. Flach, Patterns of dataset shift, 2014. URLhttps://api.semanticscholar.org/CorpusID:211733031

    2014

  8. [15]

    Zhang, B

    K. Zhang, B. Scholkopf, K. Muandet, Z. Wang, Domain adaptation under target and conditional shift, in: International Conference on Ma- chine Learning, 2013. URLhttps://api.semanticscholar.org/CorpusID:17069732

    2013

  9. [16]

    arXiv:1903.09734

    K.Azizzadenesheli, A.Liu, F.Yang, A.Anandkumar, Regularizedlearn- ing for domain adaptation under label shifts (2019). arXiv:1903.09734

    2019 arXiv

  10. [17]

    M. Wang, W. Deng, Deep visual domain adaptation: A survey, Neuro- computing 312 (2018) 135–153. doi:10.1016/j.neucom.2018.05.083. URLhttp://dx.doi.org/10.1016/j.neucom.2018.05.083

    2018 doi

  11. [18]

    S. B. David, T. Lu, T. Luu, D. Pal, Impossibility theorems for domain adaptation, in: Y. W. Teh, M. Titterington (Eds.), Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statis- tics, Vol. 9 of Proceedings of Machine Learning Research, PMLR...

    2010

  12. [19]

    J.H.Kim, B.H.Ngo, J.H.Park, J.E.Kwon, H.S.Lee, S.I.Cho, Distill- ing and refining domain-specific knowledge for semi-supervised domain adaptation., 2022

    2022

  13. [20]

    B. H. Ngo, Y. Jeong Chae, J. E. Kwon, J. Hyeon Park, S. I. Cho, Improved knowledge transfer for semi-supervised domain adap- tation via trico training strategy, in: 2023 IEEE/CVF International 33 Conference on Computer Vision (ICCV), 2023, pp. 19157–19166. doi:10.1109/ICCV5107...

    2023

  14. [21]

    Tzeng, J

    E. Tzeng, J. Hoffman, N. Zhang, K. Saenko, T. Darrell, Deep domain confusion: Maximizing for domain invariance (2014). arXiv:1412.3474

    2014 arXiv

  15. [22]

    M. Long, Y. Cao, J. Wang, M. I. Jordan, Learning trans- ferable features with deep adaptation networks (2015). doi:10.48550/ARXIV.1502.02791. URLhttps://arxiv.org/abs/1502.02791

    2015 doi arXiv

  16. [23]

    B. Sun, K. Saenko, Deep coral: Correlation alignment for deep domain adaptation (2016). arXiv:1607.01719

    2016 arXiv

  17. [24]

    C. Chen, Z. Fu, Z. Chen, S. Jin, Z. Cheng, X. Jin, X.-S. Hua, Homm: Higher-order moment matching for unsupervised domain adaptation (2019). arXiv:1912.11976

    2019 arXiv

  18. [26]

    Chuan-Xian, Z

    R. Chuan-Xian, Z. Yi-Ming, L. You-Wei, L. Meng-Xue, To- wards unsupervised domain adaptation via domain-transformer (2022). doi:10.48550/ARXIV.2202.13777. URLhttps://arxiv.org/abs/2202.13777

    2022 doi arXiv

  19. [27]

    T. Xu, W. Chen, P. Wang, F. Wang, H. Li, R. Jin, Cdtrans cross-domain transformer for unsupervised domain adaptation (2021). doi:10.48550/ARXIV.2109.06165. URLhttps://arxiv.org/abs/2109.06165

    2021 doi arXiv

  20. [28]

    J. Shen, Y. Qu, W. Zhang, Y. Yu, Wasserstein distance guided repre- sentation learning for domain adaptation (2018). arXiv:1707.01217

    2018 arXiv

  21. [29]

    Gulrajani, F

    I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin, A. Courville, Im- proved training of wasserstein gans (2017). arXiv:1704.00028

    2017 arXiv

  22. [30]

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, Y. Bengio, Generative adversarial networks (2014). arXiv:1406.2661. 34

    2014 arXiv

  23. [31]

    Isola, J.-Y

    P. Isola, J.-Y. Zhu, T. Zhou, A. A. Efros, Image-to-image translation with conditional adversarial networks (2016). doi:10.48550/ARXIV.1611.07004. URLhttps://arxiv.org/abs/1611.07004

    2016 doi arXiv

  24. [32]

    M.-Y. Liu, O. Tuzel, Coupled generative adversarial networks (2016). arXiv:1606.07536

    2016 arXiv

  25. [33]

    J.-Y. Zhu, T. Park, P. Isola, A. A. Efros, Unpaired image-to- image translation using cycle-consistent adversarial networks (2020). arXiv:1703.10593

    2020 arXiv

  26. [34]

    Zheng, Q

    W. Zheng, Q. Li, G. Zhang, P. Wan, Z. Wang, Ittr un- paired image-to-image translation with transformers (2022). doi:10.48550/ARXIV.2203.16015. URLhttps://arxiv.org/abs/2203.16015

    2022 doi arXiv

  27. [35]

    Ghifary, W

    M. Ghifary, W. B. Kleijn, M. Zhang, D. Balduzzi, Domain gener- alization for object recognition with multi-task autoencoders (2015). arXiv:1508.07680

    2015 arXiv

  28. [36]

    Caruana, Multitask learning, Machine Learning 28 (2004) 41–75

    R. Caruana, Multitask learning, Machine Learning 28 (2004) 41–75

    2004

  29. [37]

    K. He, G. Gkioxari, P. Dollár, R. Girshick, Mask r-cnn (2017). doi:10.48550/ARXIV.1703.06870. URLhttps://arxiv.org/abs/1703.06870

    2017 doi arXiv

  30. [38]

    Zhang, Q

    Y. Zhang, Q. Yang, A survey on multi-task learning (2017). doi:10.48550/ARXIV.1707.08114. URLhttps://arxiv.org/abs/1707.08114

    2017 doi arXiv

  31. [39]

    B. Lin, F. Ye, Y. Zhang, I. W. Tsang, Reasonable effectiveness of random weighting: A litmus test for multi-task learning (2022). arXiv:2111.10603

    2022 arXiv

  32. [40]

    arXiv:1711.02257

    Z.Chen, V.Badrinarayanan, C.-Y.Lee, A.Rabinovich, Gradnorm: Gra- dient normalization for adaptive loss balancing in deep multitask net- works (2018). arXiv:1711.02257

    2018 arXiv

  33. [41]

    T. Yu, S. Kumar, A. Gupta, S. Levine, K. Hausman, C. Finn, Gradient surgery for multi-task learning (2020). arXiv:2001.06782. 35

    2020 arXiv

  34. [42]

    Guangyuan, Q

    S. Guangyuan, Q. Li, W. Zhang, J. Chen, X.-M. Wu, Recon: Reduc- ing conflicting gradients from the root for multi-task learning, in: The Eleventh International Conference on Learning Representations, 2022

    2022

  35. [43]

    Z. Ren, Y. J. Lee, Cross-domain self-supervised multi-task feature learn- ing using synthetic imagery, in: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  36. [44]

    Yang, M.-C

    S.-M. Yang, M.-C. Yeh, Unsupervised multi-task domain adaptation, in: 2020 25th International Conference on Pattern Recognition (ICPR), 2021, pp. 1679–1685. doi:10.1109/ICPR48806.2021.9412458

    2020

  37. [45]

    Zhang, W

    J. Zhang, W. Li, P. Ogunbona, Unsupervised domain adaptation: A multi-task learning-based method, Knowledge-Based Systems 186 (2019) 104975. doi:https://doi.org/10.1016/j.knosys.2019.104975. URLhttps://www.sciencedirect.com/science/article/pii/ S0950705119304010

    2019

  38. [46]

    K. He, X. Zhang, S. Ren, J. Sun, Identity mappings in deep residual networks (2016). arXiv:1603.05027

    2016 arXiv

  39. [47]

    A. M. Hillas, Cerenkov Light Images of EAS Produced by Primary Gamma Rays and by Nuclei, in: 19th International Cosmic Ray Con- ference (ICRC19), Volume 3, Vol. 3 of International Cosmic Ray Con- ference, 1985, p. 445

    1985

  40. [48]

    S. Ohm, C. van Eldik, K. Egberts, gamma/hadron separation in very-high-energy gamma-ray astronomy using a multivariate analysis method, Astroparticle Physics 31 (5) (2009) 383–391. doi:https://doi.org/10.1016/j.astropartphys.2009.04.001. URLhttps://www.sciencedirect.com/scienc...

    2009 doi

  41. [49]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition (2015). doi:10.48550/ARXIV.1512.03385. URLhttps://arxiv.org/abs/1512.03385

    2015 doi arXiv

  42. [50]

    J. Hu, L. Shen, S. Albanie, G. Sun, E. Wu, Squeeze-and-excitation net- works (2019). arXiv:1709.01507. 36

    2019 arXiv

  43. [51]

    Dell’aiera, T

    M. Dell’aiera, T. Vuillaume, M. Jacquemont, A. Benoit, Deep unsupervised domain adaptation applied to the cherenkov tele- scope array large-sized telescope, in: 20th International Conference on Content-based Multimedia Indexing, CBMI 2023, ACM, 2023. doi:10.1145/3617233.361727...

    2023

  44. [52]

    Arjovsky, S

    M. Arjovsky, S. Chintala, L. Bottou, Wasserstein gan (2017). arXiv:1701.07875

    2017 arXiv

  45. [53]

    K. He, X. Chen, S. Xie, Y. Li, P. Dollár, R. Girshick, Masked autoen- coders are scalable vision learners (2021). arXiv:2111.06377

    2021 arXiv

  46. [54]

    LeCun, L

    Y. LeCun, L. Bottou, Y. Bengio, P. Haffner, Gradient-based learning appliedtodocumentrecognition, ProceedingsoftheIEEE86(11)(1998) 2278–2324

    1998

  47. [55]

    Hull, A database for handwritten text recognition research, IEEE Transactions on Pattern Analysis and Machine Intelligence 16 (5) (1994) 550–554

    J. Hull, A database for handwritten text recognition research, IEEE Transactions on Pattern Analysis and Machine Intelligence 16 (5) (1994) 550–554. doi:10.1109/34.291440

    1994 doi

  48. [56]

    Netzer, T

    Y. Netzer, T. Wang, A. Coates, A. Bissacco, B. Wu, A. Y. Ng, Reading digits in natural images with unsupervised feature learning, in: NIPS Workshop on Deep Learning and Unsupervised Feature Learning 2011, 2011. URLhttp://ufldl.stanford.edu/housenumbers/nips2011_ housenumbers.pdf

    2011

  49. [57]

    D. Heck, J. Knapp, J. N. Capdevielle, G. Schatz, T. Thouw, CORSIKA: a Monte Carlo code to simulate extensive air showers., 1998

    1998

  50. [58]

    Lopez-Coto, T

    R. Lopez-Coto, T. Vuillaume, A. Moralejo, F. Cassol, M. Nöthe, D. Mor- cuende, C. Priyadarshi, M. Bernardos, S. Nozaki, P. Gliwny, J. E. Ruiz, E. Garcia, M. Dalchenko, yrenier, L. Saha, M. Jacquemont, D. Neise, C. Alispach, R. Pillera, Andres-Baquero, J. Sitarek, aaguasca, M. ...

    2022 doi

  51. [59]

    B. Kim, A. Brill, T. Miener, D. Nieto, Q. Feng, Dl1-data-handler: Dl1 hdf5 writer, reader, and processor for iact data (Jan. 2022). 37 doi:10.5281/zenodo.5844134. URLhttps://doi.org/10.5281/zenodo.5844134

    2022 doi

  52. [60]

    Jacquemont., L

    M. Jacquemont., L. Antiga., T. Vuillaume., G. Silvestri., A. Benoit., P. Lambert., G. Maurin., Indexed operations for non-rectangular lat- tices applied to convolutional neural networks, in: Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and...

    2019 doi

  53. [61]

    CTA Observatory and Consortium, CTAO Instrument Response Func- tions - prod5 version v0.1 (Sep. 2021). doi:10.5281/zenodo.5499840. URLhttps://doi.org/10.5281/zenodo.5499840 38

    2021 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.