Pith. sign in

REVIEW 3 major objections 5 minor 40 references

UASPL: Uncertainty-Aware Self-Paced Learning with Evidential Neural Networks

T0 review · 3 major / 5 minor · reviewed 2026-07-11 · grok-4.5

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

Pith's one-line read Self-paced learning selects more reliably simple samples when model-generated evidential uncertainty is folded into the loss, not just the training error.

desk verdict Solid empirical SPL paper: internal evidential uncertainty in the selection loss is the real move; image results quietly change the selection rule, so treat generality as provisional. read the letter →

arxiv 2607.06638 v1 pith:PX2GA4PM submitted 2026-07-07 cs.LG

Yifan Zhang , Yuxin Hu , Zhuobin Hao , Xiaozhuan Gao , Lipeng Pan This is my paper · ORCID
classification cs.LG
keywords self-pacedlearningevidentialdeepuncertaintyestimationDirichletevidencesampleselectionSubjectiveLogiccurriculum
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

Self-paced learning trains models by starting with easy samples and adding harder ones, but "easy" is usually defined only by a small loss. The authors show that early low-loss samples can later become unstable, so small loss alone does not mean a sample is reliably simple for the current model. They propose UASPL: an evidential neural network that turns predictions into Dirichlet evidence and uncertainty, then builds a single sample loss that combines label fit with an uncertainty-weighted KL term. Sorting on that loss yields an interpretable order—correct low-uncertainty first, then correct high-uncertainty, then incorrect high-uncertainty, then incorrect low-uncertainty—and the same idea plugs into hard, linear, and mixture self-paced regularizers. Across many UCI tables and four image benchmarks, the method improves accuracy and stability over loss-only SPL, uncertainty-prior SPL, and several reweighting baselines, while the selection order matches the stated preference.

What carries the argument

The uncertainty-aware sample loss L_total = L_EMSE + coeff_i L_KL, with coeff_i = (1-c_i)(1-u_i)+c_i u_i, which couples prediction correctness and evidential uncertainty so that sorting samples by L_total induces the four-category easy-to-hard preference.

What would settle it

On held-out stages or datasets, check whether samples that UASPL ranks easiest in the first round keep smaller loss and higher correctness later than loss-only SPL selections; if first-round UASPL picks later become more unstable or less accurate than the baselines, the reliability claim fails.

Watch

Extended reading notes

Core claim

UASPL claims that embedding model-generated evidential uncertainty together with the label-fitting loss inside the self-paced objective selects reliably simple samples, not merely low-loss ones, and that the resulting ranking is both more accurate and more interpretable than loss-only or externally prior-based self-paced methods.

Load-bearing premise

That the hand-designed coefficient and the sorted total loss really track what will stay easy for the model later, not just how confident the current evidence looks.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes UASPL, a self-paced learning method that replaces pure loss-based sample ranking with an evidential objective. An ENN parameterizes a Dirichlet distribution over class probabilities; sample difficulty is measured by L_total = L_EMSE + coeff_i L_KL, where the adaptive coefficient coeff_i = (1-c_i)(1-u_i)+c_i u_i depends on prediction correctness and evidential uncertainty (Eqs. 6–8). Derivative analysis in §3.2 is used to argue that sorting on L_total induces an interpretable four-category easy-to-hard preference. The method is claimed to be the first SPL approach that puts model-generated evidential uncertainty inside the SPL objective, to be general across hard/linear/mixture regularizers, and to improve classification performance. Empirical support includes 25 UCI datasets (50 Monte Carlo splits), many SPL and reweighting baselines, ablations of Dirichlet modeling and KL weighting (Table 5), generality tables (Tables 3–4), and additional image-benchmark results under deep backbones.

Significance. If the reliability claim holds, the paper offers a concrete and reusable way to inject model-internal predictive reliability into SPL without external priors, with a clear selection story and code release. Strengths include a broad UCI evaluation with many baselines (including adapted uncertainty-SPL variants), ablations that isolate Dirichlet modeling and the adaptive KL coefficient (Table 5), generality checks across three regularizers, and supplementary noise/pretraining/Wilcoxon analyses. The contribution is incremental rather than foundational—it sits at the intersection of EDL and SPL—but it is practically useful if the same reliability mechanism is shown to drive gains under both tabular and deep-image settings.

major comments (3)
  1. The load-bearing reliability mechanism analyzed in §3.2 is pure ranking by L_total with the adaptive coeff_i of Eqs. 6–8. In §4.5.2 the image experiments abandon that operational definition: they add a correct-evidence regularizer and replace selection by the stage-dependent hybrid score_i = (1-r_t) Norm(L_total) + r_t Norm(1-u_i) (Eq. 18). Consequently, the CIFAR-10/FashionMNIST/MNIST/SVHN results (Table 6) do not test the same selection rule that Contributions (1)–(2) and the four-category preference claim. Either re-run image experiments under pure L_total ranking (or report both), or explicitly restate the image protocol as a separate practical variant and limit the theoretical reliability claim to the tabular setting.
  2. The interpretability argument in §3.2 is partly by construction: L_total is designed so that sorting yields correct/low-u → correct/high-u → incorrect/high-u → incorrect/low-u, and Fig. 4 then visualizes that order. That is useful design analysis, but it does not by itself establish that the selected samples remain ‘reliably easy’ for later stages. The MRLV diagnostic (Figs. 1 and 3) is only a proxy. Strengthen the claim with an independent check—e.g., hold-out difficulty labels, later-stage loss trajectories stratified by the four categories, or a controlled comparison against a non-adaptive coeff—so that reliability is not identified solely with the designed ranking.
  3. Table 6 shows UASPL is competitive on images but not uniformly superior (e.g., CIFAR-10 mean accuracy trails MW-Net; several methods are near saturation on MNIST). Given that the image protocol already differs from the analyzed objective, the paper should either (i) provide matched ablations of Eq. 18 and the correct-evidence term under deep backbones, or (ii) temper the generality claim so that ‘outperforms other SPL methods … and generality’ is not read as applying equally to the deep-image regime.
minor comments (5)
  1. Algorithm 1 is hard to follow: the first for-loop over epochs appears to only compute α/S, while selection and updates sit outside a clear outer training loop; clarify the pacing schedule and when λ is updated.
  2. Notation for Dirichlet parameters and evidence (α_ij, e_ij, S_i) is introduced more than once with slight redundancy between §2.1 and §3.1; a single consistent block would help.
  3. Table 1’s checkmarks for ‘Internal / Dynamic / Selection criterion’ are useful but slightly overloaded; a short footnote defining each column would reduce ambiguity.
  4. Several detailed results are deferred to the Supplementary Material (F1/Precision/Recall tables, noise and Wilcoxon tests). Ensure the main text states the key numerical takeaways so the paper is self-contained for readers who do not open the appendix.
  5. Minor prose issues: ‘this article’ vs ‘this paper’, occasional long sentences in §3.2, and arXiv ID / date formatting should be cleaned for journal submission.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild self-definitional design of the selection preference; empirical performance claims rest on external benchmarks and are not forced by construction.

  1. self definitional [§3.1–3.2, Eqs. 6–8; Contribution (2); Abstract]
    "coeffi = (1−ci)(1−ui)+ci ui ... Based on the above analysis, the sample-selection preference induced by UASPL can be revealed by sorting L(i)total in ascending order, yielding the following easier-to-harder sequence: • correct prediction with low uncertainty • correct prediction with high uncertainty • incorrect prediction with high uncertainty • incorrect prediction with low uncertainty ... this loss function couples a sample selection preference, thereby ensuring the interpretability of the sample selection process."

    The coefficient is defined so that, among correct samples, lower u shrinks the KL term (smaller L_total), and among incorrect samples, higher u shrinks the KL term. Sorting L_total therefore produces the four-category order by design of coeff_i (plus the usual EMSE gap between correct and incorrect). The paper then presents that order as an ‘induced’ interpretable preference that ‘ensures’ reliability-aware selection—i.e., the claimed preference is largely the definition of the loss, not an independent derivation.

full rationale

UASPL’s central empirical claims (accuracy/F1/etc. vs. SPL, reweighting, and uncertainty-adapted baselines on held-out UCI and image data; Tables 2–6) are evaluated against external methods and test metrics and do not reduce to fitted inputs or self-citation uniqueness theorems. The only circularity-adjacent step is interpretability Contribution (2) / §3.2: the adaptive coefficient coeff_i is hand-designed so that sorting L_total = L_EMSE + coeff_i L_KL yields the four-category order the authors then present as the induced easy-to-hard preference. That ordering is largely by construction of coeff (and the usual fact that correct predictions have smaller EMSE), not an independent first-principles prediction. This is mild design circularity of the narrative, not a tautology that forces the reported accuracy gains. No load-bearing self-citation uniqueness chain or fitted-parameter-as-prediction pattern is present. The image hybrid score (Eq. 18) is a consistency/scope issue, not circularity of the derivation.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The central claim rests on standard SPL and EDL machinery plus one paper-specific adaptive coefficient and selection schedule. Free parameters are mostly training/pacing choices rather than physics-style fitted constants. Invented content is the UASPL loss/preference design, not a new physical entity.

free parameters (4)
  • SPL age/pacing schedule λ and stage fractions = 6 stages: 25% then +15% each stage
    Hard regularizer age parameter and fixed six-stage schedule (start 25%, then +15%) control which samples enter training; results depend on this hand-chosen curriculum.
  • Pretraining duration = 20 epochs (tabular), 5 epochs (image)
    Model is pretrained with cross-entropy before SPL (20 epochs tabular, 5 image); sensitivity is only partially checked in the supplement.
  • Adaptive KL coefficient form coeff_i = coeff_i=(1-c_i)(1-u_i)+c_i u_i
    The mapping from correctness c_i and uncertainty u_i to KL weight is a design choice, not derived from a uniqueness theorem; ablations compare fixed/annealed alternatives but the functional form is free.
  • Image hybrid selection score weight r_t and correct-evidence regularization = r_t=(t+1)/T; correct-evidence term from cited EDL work
    Image experiments alter the selection criterion and add an extra regularizer not present in the main tabular objective; these are additional free design knobs.
assumptions (5)
  • domain assumption Predictive uncertainty is adequately represented by Dirichlet evidence parameters under Subjective Logic (u=K/S).
    Imported from evidential deep learning (§2.1); the reliability claim depends on this uncertainty being meaningful for sample difficulty.
  • domain assumption Self-paced learning with sample weights v_i and regularizer g(v;λ) is a valid progressive training framework.
    Standard SPL objective (Eq. 5) is taken as given; UASPL only changes the per-sample loss used to set difficulty.
  • domain assumption EMSE plus a KL penalty on non-target evidence is an appropriate evidential classification loss.
    Uses the common EDL loss family (Eqs. 1–3) as the base of L_total.
  • ad hoc to paper During training, the correctness indicator c_i from argmax prediction vs label is a legitimate signal for reweighting the KL term.
    Labels are available, but using instantaneous correctness to define sample reliability is a paper-specific modeling choice (Eq. 8).
  • ad hoc to paper Mean relative loss variation of first-round samples is a diagnostic of whether those samples were truly easy.
    Motivation in Fig. 1 and later MRLV comparisons depends on this diagnostic; it is not a standard external benchmark.
invented entities (2)
  • Uncertainty-aware SPL loss L_total with adaptive coeff_i
    purpose: Fuse label fit and evidential uncertainty into one sample-difficulty score for SPL selection.
    Core proposed object (Eqs. 6–8); evaluated only inside this paper's experiments, though built from known EDL terms.
  • Four-category sample-selection preference (correct/low-u → correct/high-u → incorrect/high-u → incorrect/low-u)
    purpose: Provide an interpretable easy-to-hard ordering induced by the loss.
    Presented as a consequence of the designed loss (§3.2); independent external validation of this taxonomy as true difficulty is limited.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UASPL: Uncertainty-Aware Self-Paced Learning with Evidential Neural Networks." pith.science (2026). https://pith.science/paper/PX2GA4PM

@misc{pith2026260706638,
  author       = {Pith},
  title        = {Pith review of: UASPL: Uncertainty-Aware Self-Paced Learning with Evidential Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PX2GA4PM}},
  note         = {Machine review of arXiv:2607.06638}
}
read the original abstract

Self-paced learning (SPL) is an effective learning paradigm that simulates the human learning process by progressing from easy to difficult samples based on the value of the loss function during the learning process. It has shown great potential in improving model performance and training efficiency. However, the prediction results of samples with smaller loss values are not necessarily reliable, indicating that such samples are not always simple samples for the model. Hence, this article proposes an uncertainty-aware self-paced learning based on evidential neural networks, termed UASPL, which integrates predictive reliability into sample selection through a general loss function within the Subjective Logic framework. This loss function incorporates uncertainty estimation and can be extended to different variants of SPL. Moreover, this loss function couples a sample selection preference, thereby ensuring the interpretability of the sample selection process. Finally, the experimental results on multiple datasets show that UASPL outperforms other SPL methods in terms of classification performance, interpretability, and generality. The source code is available at: https://github.com/treelife979/UASPL.

Figures

Figures reproduced from arXiv: 2607.06638 by the authors.

Figure 1
Figure 1. Mean relative loss variation across rounds on four datasets. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Framework of UASPL and comparison with prior SPL methods. Traditional SPL estimates sample [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Comparison of mean relative loss variation across rounds for UASPL and representative baselines [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Prediction results and uncertainty of UASPL with loss ascending sorting on the Transfusion dataset. [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    M. P. Kumar, B. Packer, D. Koller, Self-paced learning for latent variable models, in: Advances in Neural Information Processing Systems, V ol. 23, Curran Asso- ciates, Inc., 2010, pp. 1189–1197

    work page 2010

  2. [2]

    Bengio, J

    Y . Bengio, J. Louradour, R. Collobert, J. Weston, Curriculum learning, in: Pro- ceedings of the 26th Annual International Conference on Machine Learning, ACM, 2009, pp. 41–48.doi:10.1145/1553374.1553380

    work page doi:10.1145/1553374.1553380 2009

  3. [3]

    W. Li, H. Chen, T. Li, J. Wan, B. Sang, Unsupervised feature selection via self- paced learning and low-redundant regularization, Knowledge-Based Systems 240 (2022) 108150.doi:10.1016/j.knosys.2022.108150

    work page doi:10.1016/j.knosys.2022.108150 2022

  4. [4]

    Y . Yang, H. Chen, J. Hu, X. Guo, T. Ma, Advancing Brain Imaging Analysis Step-by-Step via Progressive Self-paced Learning, Springer Nature Switzerland, 2024, pp. 58–68

    work page 2024

  5. [5]

    D. Zhao, Y . S. Koh, G. Dobbie, H. Hu, P. Fournier-Viger, Symmetric self-paced learning for domain generalization, Proceedings of the AAAI Conference on Ar- tificial Intelligence 38 (15) (2024) 16961–16969.doi:10.1609/aaai.v38i15. 29639

    work page doi:10.1609/aaai.v38i15 2024

  6. [6]

    Jiang, D

    L. Jiang, D. Meng, T. Mitamura, A. G. Hauptmann, Easy samples first: Self- paced reranking for zero-example multimedia search, in: Proceedings of the 22nd ACM international conference on Multimedia, MM ’14, ACM, 2014, pp. 547– 556.doi:10.1145/2647868.2654918

    work page doi:10.1145/2647868.2654918 2014

  7. [7]

    Q. Zhao, D. Meng, L. Jiang, Q. Xie, Z. Xu, A. Hauptmann, Self-paced learning for matrix factorization, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 29, 2015, pp. 3196–3202.doi:10.1609/aaai.v29i1.9584. 30

    work page doi:10.1609/aaai.v29i1.9584 2015

  8. [8]

    C. Li, F. Wei, J. Yan, X. Zhang, Q. Liu, H. Zha, A self-paced regularization frame- work for multilabel learning, IEEE Transactions on Neural Networks and Learn- ing Systems 29 (6) (2018) 2660–2666.doi:10.1109/tnnls.2017.2697767

    work page doi:10.1109/tnnls.2017.2697767 2018

Show all 40 references
  1. [9]

    Jiang, D

    L. Jiang, D. Meng, Q. Zhao, S. Shan, A. Hauptmann, Self-paced curriculum learn- ing, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 29, 2015, pp. 2694–2700.doi:10.1609/aaai.v29i1.9608

    2015 doi

  2. [10]

    GUO, Self-paced learning with statistics uncertainty prior, IEICE Transac- tions on Information and Systems E101.D (3) (2018) 812–816.doi:10.1587/ transinf.2017edl8169

    L. GUO, Self-paced learning with statistics uncertainty prior, IEICE Transac- tions on Information and Systems E101.D (3) (2018) 812–816.doi:10.1587/ transinf.2017edl8169

    2018

  3. [11]

    Zhang, D

    S. Zhang, D. Han, J. Dezert, Y . Yang, Weighted self-paced learning with belief functions, Expert Systems with Applications 255 (2024) 124535.doi:10.1016/ j.eswa.2024.124535

    2024

  4. [12]

    Chang, E

    H.-S. Chang, E. Learned-Miller, A. McCallum, Active bias: Training more ac- curate neural networks by emphasizing high variance samples, in: Advances in Neural Information Processing Systems, V ol. 30, Curran Associates, Inc., 2017, pp. 1002–1012

    2017

  5. [13]

    Islam, L

    M. Islam, L. Seenivasan, S. P. Sharan, V . K. Viekash, B. Gupta, B. Glocker, H. Ren, Paced-curriculum distillation with prediction and label uncertainty for image segmentation, International Journal of Computer Assisted Radiology and Surgery 18 (10) (2023) 1875–1883.doi:10.100...

    2023 doi

  6. [14]

    Q. Feng, Y . Liu, H. Schuetze, Your pretrained model tells the difficulty it- self: A self-adaptive curriculum learning paradigm for natural language un- derstanding, in: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (V olume 4: Studen...

    2025 doi

  7. [15]

    Y . Gal, Z. Ghahramani, Dropout as a bayesian approximation: Representing model uncertainty in deep learning, in: Proceedings of The 33rd International Conference on Machine Learning, V ol. 48 of Proceedings of Machine Learning Research, PMLR, New York, New York, USA, 2016, pp...

    2016

  8. [16]

    Y . Gal, R. Islam, Z. Ghahramani, Deep Bayesian active learning with image data, in: Proceedings of the 34th International Conference on Machine Learn- ing, V ol. 70 of Proceedings of Machine Learning Research, PMLR, 2017, pp. 1183–1192

    2017

  9. [17]

    K. Kim, D. Park, K. I. Kim, S. Y . Chun, Task-aware variational adversarial active learning, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 8166–8175

    2021

  10. [18]

    J. O. Woo, Active learning in bayesian neural networks with balanced entropy learning principle, in: The Eleventh International Conference on Learning Rep- resentations, 2023

    2023

  11. [19]

    Sensoy, L

    M. Sensoy, L. Kaplan, M. Kandemir, Evidential deep learning to quantify clas- sification uncertainty, in: Advances in Neural Information Processing Systems, V ol. 31, Curran Associates, Inc., 2018, pp. 3179–3189

    2018

  12. [20]

    Zhang, P

    Q. Zhang, P. Zhang, T. Li, Information fusion for large-scale multi-source data based on the dempster-shafer evidence theory, Information Fusion 115 (2025) 102754.doi:10.1016/j.inffus.2024.102754

    2025 doi

  13. [21]

    Y . Dong, N. Jiang, R. Zhou, C. Zhu, L. Cao, T. Liu, Y . Xu, X. Li, A novel multi-criteria conflict evidence combination method and its application to pattern recognition, Information Fusion 108 (2024) 102346.doi:10.1016/j.inffus. 2024.102346. 32

    2024 doi

  14. [22]

    Zhang, Y

    Z. Zhang, Y . Zhang, H. Tian, A. Martin, Z. Liu, W. Ding, A survey of eviden- tial clustering: Definitions, methods, and applications, Information Fusion 115 (2025) 102736.doi:10.1016/j.inffus.2024.102736

    2025 doi

  15. [23]

    M. Chen, J. Gao, C. Xu, Revisiting essential and nonessential settings of eviden- tial deep learning, IEEE Transactions on Pattern Analysis and Machine Intelli- gence 47 (10) (2025) 8658–8673.doi:10.1109/tpami.2025.3583410

    2025 doi

  16. [24]

    P. Wang, S. Zhang, R. Chen, J. Chen, Y . Fan, L. Ning, Y . Cao, Regularized evi- dential neural networks for deep active learning, Pattern Recognition 173 (2026) 112836.doi:10.1016/j.patcog.2025.112836

    2026 doi

  17. [25]

    C. Wang, X. Wang, J. Zhang, L. Zhang, X. Bai, X. Ning, J. Zhou, E. Hancock, Uncertainty estimation for stereo matching based on evidential deep learning, Pat- tern Recognition 124 (2022) 108498.doi:10.1016/j.patcog.2021.108498

    2022 doi

  18. [26]

    J. Gao, M. Chen, L. Xiang, C. Xu, A comprehensive survey on evidential deep learning and its applications, IEEE Transactions on Pattern Analysis and Machine Intelligence 48 (3) (2026) 2118–2138.doi:10.1109/tpami.2025.3625258

    2026 doi

  19. [27]

    D. Meng, Q. Zhao, L. Jiang, A theoretical understanding of self-paced learn- ing, Information Sciences 414 (2017) 319–328.doi:10.1016/j.ins.2017. 05.043

    2017 doi

  20. [28]

    Z. Kang, H. Liu, J. Li, X. Zhu, L. Tian, Self-paced principal component analy- sis, Pattern Recognition 142 (2023) 109692.doi:10.1016/j.patcog.2023. 109692

    2023 doi

  21. [29]

    J. Yan, W. Wei, X. Guo, C. Dang, J. Liang, A bi-level metric learning framework via self-paced learning weighting, Pattern Recognition 139 (2023) 109446.doi: 10.1016/j.patcog.2023.109446. 33

    2023 doi

  22. [30]

    B. Kang, S. Du, D. Liang, F. Wu, X. Li, Fine-grained recognition via submod- ular optimization regulated progressive training, Pattern Recognition 156 (2024) 110849.doi:10.1016/j.patcog.2024.110849

    2024 doi

  23. [31]

    Poyser, T

    M. Poyser, T. P. Breckon, Dds-nas: Dynamic data selection within neural ar- chitecture search via on-line hard example mining applied to image classifica- tion, Pattern Recognition 169 (2026) 111948.doi:10.1016/j.patcog.2025. 111948

    2026 doi

  24. [32]

    H. Li, M. Gong, Self-paced convolutional neural networks, in: Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, IJCAI- 2017, International Joint Conferences on Artificial Intelligence Organization, 2017, pp. 2110–2116.doi:10.24963/ijcai...

    2017 doi

  25. [33]

    H. Li, J. Li, Y . Zhao, M. Gong, Y . Zhang, T. Liu, Cost-sensitive self-paced learn- ing with adaptive regularization for classification of image time series, IEEE Jour- nal of Selected Topics in Applied Earth Observations and Remote Sensing 14 (2021) 11713–11727.doi:10.1109/j...

    2021 doi

  26. [34]

    Y . Xu, R. Liu, T. Xiahou, X. Li, Y . Liu, Interactive cost-based reliability con- sensus reaching models for multisource imprecise information calibration of multistate systems, IEEE Transactions on Reliability 74 (4) (2025) 4842–4857. doi:10.1109/tr.2025.3562084

    2025 doi

  27. [35]

    R. Liu, T. Xiahou, Y . Liu, Multisource imprecise information calibration for relia- bility assessment of multistate systems: A consensus reaching perspective, IEEE Transactions on Reliability 74 (1) (2025) 2226–2240.doi:10.1109/tr.2024. 3393985

    2025 doi

  28. [36]

    J. Shu, Q. Xie, L. Yi, Q. Zhao, S. Zhou, Z. Xu, D. Meng, Meta-weight-net: Learn- ing an explicit mapping for sample weighting, in: Advances in Neural Informa- tion Processing Systems, V ol. 32, Curran Associates, Inc., 2019, pp. 1917–1928. 34

    2019

  29. [37]

    X. Li, W. Shen, D. Charles, Tedl: A two-stage evidential deep learning method for classification uncertainty quantification, in: Proceedings of the Workshop on Deep Learning for Search and Recommendation (DL4SR 2022), V ol. 3317 of CEUR Workshop Proceedings, 2022

    2022

  30. [38]

    Q. Zhao, J. Shu, X. Yuan, Z. Liu, D. Meng, A probabilistic formulation for meta- weight-net, IEEE Transactions on Neural Networks and Learning Systems 34 (3) (2023) 1194–1208.doi:10.1109/TNNLS.2021.3105104

    2023 doi

  31. [39]

    J. Shu, X. Yuan, D. Meng, Z. Xu, Cmw-net: Learning a class-aware sam- ple weighting mapping for robust deep learning, IEEE Transactions on Pat- tern Analysis and Machine Intelligence 45 (10) (2023) 11521–11539.doi: 10.1109/TPAMI.2023.3271451

    2023 doi

  32. [40]

    D. S. Pandey, H. Choi, Q. Yu, Generalized regularized evidential deep learning models: Theory and comprehensive evaluation, IEEE Transactions on Pattern Analysis and Machine Intelligence 48 (6) (2026) 6865–6879.doi:10.1109/ tpami.2026.3660699. 35

    2026

Pith tools

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