Pith. sign in

REVIEW 2 major objections 2 minor 115 references

Bandwidth Selection in Kernel Density Estimation for Model Calibration

T0 review · 2 major / 2 minor · reviewed 2026-06-30 · grok-4.3

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

Pith's one-line read Aligning KDE-reconstructed risk with empirical risk selects bandwidths that minimize calibration estimation bias

desk verdict Risk Alignment gives a calibration-focused bandwidth selector for KDE, but the bias-minimization theory is asserted without visible derivation. read the letter →

arxiv 2606.29925 v1 pith:6MSR6JEQ submitted 2026-06-29 cs.LG

Han Zhou , Teodora Popordanoska , Matthew Blaschko This is my paper · ORCID
classification cs.LG
keywords kerneldensityestimationbandwidthselectionmodelcalibrationriskalignmenterrordeeplearninguncertainty
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

This paper introduces Risk Alignment, a method to choose the kernel bandwidth when using KDE to measure how well-calibrated a model's predictions are. It optimizes the bandwidth so the risk value reconstructed from the smoothed density matches the risk computed directly from the data points. The authors prove theoretically that this matching step reduces bias in the resulting calibration metric over the full data distribution. The approach is presented as a general criterion that works for several calibration scores, including the canonical calibration error case where standard choices often fail. Reliable bandwidth selection matters because deep learning models are used in settings where accurate uncertainty reporting affects decisions.

What carries the argument

Risk Alignment, the optimization framework that selects kernel bandwidth by equating KDE-reconstructed risk to empirical risk

What would settle it

A dataset or synthetic case where the bandwidth chosen by Risk Alignment produces higher bias in a calibration metric than the bandwidth chosen by maximum likelihood estimation.

Watch

Extended reading notes

Core claim

Risk Alignment determines the optimal bandwidth for KDE-based calibration by aligning the reconstructed risk with the empirical risk. This alignment is shown to minimize calibration estimation bias across the data distribution and serves as a principled selection criterion for various metrics including canonical calibration error.

Load-bearing premise

That aligning the KDE-reconstructed risk with the empirical risk produces the bandwidth that minimizes bias in the calibration estimate.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper introduces Risk Alignment (RA), an optimization framework for selecting the kernel bandwidth in KDE-based calibration error estimation. By aligning the KDE-reconstructed risk with the empirical risk, the authors claim a theoretical guarantee that this choice minimizes estimation bias across the data distribution for multiple calibration metrics, including the canonical calibration error. They further report that RA outperforms standard bandwidth selectors such as MLE in experiments across architectures and datasets.

Significance. If the central theoretical claim is correct, the work supplies a task-specific, bias-minimizing criterion for KDE bandwidth selection that is directly relevant to reliable calibration assessment in deployed models. This addresses a practical weakness of KDE (suboptimal bandwidths from likelihood-based criteria) and extends to the non-trivial case of canonical calibration error. The empirical results, if reproducible, would strengthen the case for adopting RA in calibration pipelines.

major comments (2)
  1. [Theoretical section (derivation of bias minimization)] The theoretical demonstration that RA minimizes bias for canonical calibration error (the integral of |p − E[y|p]|) is not shown to follow directly from risk alignment. The derivation must explicitly cancel or bound the leading bias term in the calibration estimator; without that step or the required assumptions on the conditional density and kernel support, the claim that alignment controls this integral remains unverified.
  2. [Theoretical demonstration] The paper states that RA is 'applicable to various metrics, including the challenging case of canonical calibration error,' yet the provided derivation details do not address whether an L2-style risk discrepancy suffices for the absolute-deviation integral without further conditions. This is load-bearing for the central claim.
minor comments (2)
  1. Notation for the RA objective and the KDE-reconstructed risk should be introduced with explicit equations before the theoretical argument.
  2. The experimental section would benefit from reporting the precise quantitative improvement (e.g., reduction in calibration error) and an ablation isolating the contribution of the alignment term versus other factors.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on the theoretical claims. We address each major comment below and will strengthen the derivation in the revised manuscript.

read point-by-point responses

  1. Referee: [Theoretical section (derivation of bias minimization)] The theoretical demonstration that RA minimizes bias for canonical calibration error (the integral of |p − E[y|p]|) is not shown to follow directly from risk alignment. The derivation must explicitly cancel or bound the leading bias term in the calibration estimator; without that step or the required assumptions on the conditional density and kernel support, the claim that alignment controls this integral remains unverified.

    Authors: We agree that the current theoretical section does not explicitly derive how the L2 risk alignment cancels or bounds the leading bias term specifically for the canonical calibration error integral. In the revision we will add a dedicated subsection that starts from the risk alignment objective, expands the bias of the KDE-based estimator for the absolute-deviation integral, and states the required assumptions on the conditional density and kernel support under which the alignment controls this term. revision: yes

  2. Referee: [Theoretical demonstration] The paper states that RA is 'applicable to various metrics, including the challenging case of canonical calibration error,' yet the provided derivation details do not address whether an L2-style risk discrepancy suffices for the absolute-deviation integral without further conditions. This is load-bearing for the central claim.

    Authors: We acknowledge that the manuscript does not yet spell out the additional conditions needed for the L2 risk discrepancy to control the absolute-deviation integral. The revision will explicitly list these conditions (including kernel support and smoothness of the conditional expectation) and show the step that bridges the L2 alignment to the canonical metric, thereby verifying the applicability claim. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: theoretical claim presented as independent derivation without visible reduction to inputs or self-citation chains in abstract.

full rationale

The provided abstract states a theoretical demonstration that risk alignment minimizes calibration estimation bias, but contains no equations, no fitting procedures, and no citations. Without explicit derivation steps or self-referential definitions in the visible text, the central claim cannot be shown to reduce by construction to its inputs. The reader's note correctly flags that circularity is not visible from the abstract alone. Per hard rules, no circularity is claimed absent quotable reduction; the derivation is treated as self-contained pending full equations.

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

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the central claim rests on an unshown theoretical demonstration and unspecified experimental setup.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bandwidth Selection in Kernel Density Estimation for Model Calibration." pith.science (2026). https://pith.science/paper/6MSR6JEQ

@misc{pith2026260629925,
  author       = {Pith},
  title        = {Pith review of: Bandwidth Selection in Kernel Density Estimation for Model Calibration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6MSR6JEQ}},
  note         = {Machine review of arXiv:2606.29925}
}
read the original abstract

As deep learning models are increasingly deployed in high-stakes applications, providing well-calibrated uncertainty estimates has become as critical as achieving high predictive accuracy. While Kernel Density Estimation (KDE) has emerged as a smooth and continuous alternative to traditional binning for quantifying miscalibration, its reliability is heavily dependent on the choice of the kernel bandwidth. Standard selection techniques, such as Maximum Likelihood Estimation (MLE), often fail to produce optimal bandwidths for calibration tasks. In this work, we introduce Risk Alignment (RA), a novel optimization framework that determines the optimal bandwidth by aligning KDE-reconstructed risk with empirical risk. We theoretically demonstrate that this alignment minimizes calibration estimation bias across the data distribution, establishing a principled bandwidth selection criterion applicable to various metrics, including the challenging case of canonical calibration error. Extensive experiments across multiple architectures and datasets show that RA consistently outperforms standard bandwidth selection methods, yielding more reliable calibration assessments.

Figures

Figures reproduced from arXiv: 2606.29925 by the authors.

Figure 1
Figure 1. Bandwidth selection behavior on synthetic data. 1a and 1b illustrates the results of Rˆi using the bandwidth h selected by MLE or RA objective. While MLE is designed for density recovery, it tends to favor pathologically smaller bandwidths compared with RA objective. This preference leads to the behavior shown in the reliability plots (1a, 1b and 1e), where the MLE-selected bandwidth causes the estimated bias of Rˆi… view at source ↗
Figure 2
Figure 2. (Synthetic data) Finite sample estimator of CWCE2 2 : mean and 95% confidence interval computation across all subsamples of synthetic data [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. (Synthetic data) Finite sample estimator of CE2 2 : mean and 95% confidence interval computation across all subsamples of synthetic data [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (16 more)
Figure 4
Figure 4. Figure 4: (Synthetic data (t1 = 1.0, t2 = 0.8)) Finite sample estimator of CWCEKL: mean and 95% confidence interval computation across all subsamples of synthetic data [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: (Synthetic data (t1 = 1.0, t2 = 0.8)) Finite sample estimator of CE2 KL: mean and 95% confidence interval computation across all subsamples of synthetic data [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: (CIFAR10) Finite sample estimator of CWCE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: (CIFAR100) Finite sample estimator of CWCE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: (Amazon) Finite sample estimator of CWCE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: (ImageNet) Finite sample estimator of CWCE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: (CIFAR10) Finite sample estimator of CWCE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: (CIFAR100) Finite sample estimator of CWCE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: (Amazon) Finite sample estimator of CWCE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]
Figure 13
Figure 13. Figure 13: (ImageNet) Finite sample estimator of CWCE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]
Figure 14
Figure 14. Figure 14: (CIFAR10) Finite sample estimator of CE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: (CIFAR100) Finite sample estimator of CE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 16
Figure 16. Figure 16: (Amazon) Finite sample estimator of CE2 2 : mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p021_16.png]
Figure 17
Figure 17. Figure 17: (CIFAR10) Finite sample estimator of CE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p021_17.png]
Figure 18
Figure 18. Figure 18: (CIFAR100) Finite sample estimator of CE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p021_18.png]
Figure 19
Figure 19. Figure 19: (Amazon) Finite sample estimator of CE2 KL: mean and 95% confidence interval computation across all subsamples [PITH_FULL_IMAGE:figures/full_fig_p022_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

115 extracted references · 7 canonical work pages

  1. [1]

    1999 , publisher=

    Elements of information theory , author=. 1999 , publisher=

    1999

  2. [2]

    Advances in neural information processing systems , volume=

    Calibrating deep neural networks using focal loss , author=. Advances in neural information processing systems , volume=

  3. [3]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Obtaining well calibrated probabilities using bayesian binning , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  4. [4]

    Transactions on Machine Learning Research , year=

    Optimizing Estimators of Squared Calibration Errors in Classification , author=. Transactions on Machine Learning Research , year=

  5. [5]

    1999 , publisher=

    Interpolation of spatial data: some theory for kriging , author=. 1999 , publisher=

    1999

  6. [6]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops , pages=

    Revisiting the evaluation of uncertainty estimation and its application to explore model complexity-uncertainty trade-off , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops , pages=

  7. [7]

    The Journal of Machine Learning Research , volume=

    Gaussian processes for machine learning (GPML) toolbox , author=. The Journal of Machine Learning Research , volume=. 2010 , publisher=

    2010

  8. [8]

    The 22nd international conference on artificial intelligence and statistics , pages=

    Evaluating model calibration in classification , author=. The 22nd international conference on artificial intelligence and statistics , pages=. 2019 , organization=

    2019

Show all 115 references
  1. [9]

    International Conference on Learning Representations , year=

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale , author=. International Conference on Learning Representations , year=

  2. [10]

    The Eleventh International Conference on Learning Representations , year=

    Moderate coreset: A universal method of data selection for real-world data-efficient deep learning , author=. The Eleventh International Conference on Learning Representations , year=

  3. [11]

    2024 , publisher =

    Kentaro Yoshioka , title =. 2024 , publisher =

    2024

  4. [12]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Swin transformer: Hierarchical vision transformer using shifted windows , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  5. [13]

    Computer Vision--ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11--14, 2016, Proceedings, Part IV 14 , pages=

    Identity mappings in deep residual networks , author=. Computer Vision--ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11--14, 2016, Proceedings, Part IV 14 , pages=. 2016 , organization=

    2016

  6. [14]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Rethinking the inception architecture for computer vision , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  7. [15]

    arXiv e-prints , pages=

    DeepSeek-V3 Technical Report , author=. arXiv e-prints , pages=

  8. [16]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    arXiv

  9. [17]

    European Journal of Cancer , volume=

    Skin cancer classification via convolutional neural networks: systematic review of studies involving human experts , author=. European Journal of Cancer , volume=. 2021 , publisher=

    2021

  10. [18]

    International Conference on Learning Representations , year=

    Top-label calibration and multiclass-to-binary reductions , author=. International Conference on Learning Representations , year=

  11. [19]

    2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA) , pages=

    Class-wise and reduced calibration methods , author=. 2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA) , pages=. 2022 , organization=

    2022

  12. [20]

    Advances in neural information processing systems , volume=

    Verified uncertainty calibration , author=. Advances in neural information processing systems , volume=

  13. [21]

    International Conference on Learning Representations, Virtual conference, May 3-May 7, 2021 , pages=

    Calibration tests beyond classification , author=. International Conference on Learning Representations, Virtual conference, May 3-May 7, 2021 , pages=. 2021 , organization=

    2021

  14. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    On focal loss for class-posterior probability estimation: A theoretical perspective , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  15. [23]

    Advances in neural information processing systems , volume=

    Differentiable ranking and sorting using optimal transport , author=. Advances in neural information processing systems , volume=

  16. [24]

    Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part I 15 , pages=

    Novel decompositions of proper scoring rules for classification: Score adjustment as precursor to calibration , author=. Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part I 15 , ...

    2015

  17. [25]

    Reliability, Sufficiency, and the Decomposition of Proper Scores , journal =

    Jochen Br. Reliability, Sufficiency, and the Decomposition of Proper Scores , journal =. 2008 , volume =

    2008

  18. [26]

    Advances in Neural Information Processing Systems , volume=

    Better uncertainty calibration via proper scores for classification and beyond , author=. Advances in Neural Information Processing Systems , volume=

  19. [27]

    Pattern Recognition Letters , year=

    Geometric insights into focal loss: Reducing curvature for enhanced model calibration , author=. Pattern Recognition Letters , year=

  20. [28]

    , author=

    Existence and uniqueness of proper scoring rules. , author=. J. Mach. Learn. Res. , volume=

  21. [29]

    International Conference on Machine Learning , YEAR =

    Meta-Cal: Well-controlled Post-hoc Calibration by Ranking , AUTHOR =. International Conference on Machine Learning , YEAR =

  22. [30]

    International conference on machine learning , pages=

    On calibration of modern neural networks , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  23. [31]

    2009 , author =

    Kumaraswamy’s distribution: A beta-type distribution with some tractability advantages , journal =. 2009 , author =

    2009

  24. [32]

    1967 , issn =

    The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming , journal =. 1967 , issn =. doi:https://doi.org/10.1016/0041-5553(67)90040-7 , url =

    1967 doi

  25. [33]

    GitHub repository , doi =

    Ross Wightman , title =. GitHub repository , doi =. 2019 , publisher =

    2019

  26. [34]

    Ovcharov , title =

    Evgeni Y. Ovcharov , title =. Bernoulli , number =. 2018 , doi =

    2018

  27. [35]

    Better Uncertainty Calibration via Proper Scores for Classification and Beyond , volume =

    Gruber, Sebastian and Buettner, Florian , booktitle =. Better Uncertainty Calibration via Proper Scores for Classification and Beyond , volume =

  28. [36]

    Advances in Neural Information Processing Systems , volume=

    A consistent and differentiable lp canonical calibration error estimator , author=. Advances in Neural Information Processing Systems , volume=

  29. [37]

    Journal of Multivariate Analysis , volume=

    Asymptotic properties of Dirichlet kernel density estimators , author=. Journal of Multivariate Analysis , volume=. 2022 , publisher=

    2022

  30. [38]

    Procedings of the British Machine Vision Conference 2016 , pages=

    Wide Residual Networks , author=. Procedings of the British Machine Vision Conference 2016 , pages=. 2016 , organization=

    2016

  31. [39]

    Nature Medicine , volume=

    Enhancing the reliability and accuracy of AI-enabled diagnosis via complementarity-driven deferral to clinicians , author=. Nature Medicine , volume=. 2023 , publisher=

    2023

  32. [40]

    Advances in Neural Information Processing Systems , volume=

    Soft calibration objectives for neural networks , author=. Advances in Neural Information Processing Systems , volume=

  33. [41]

    Proceedings of the 35th International Conference on Machine Learning , pages=

    Trainable calibration measures for neural networks from kernel mean embeddings , author=. Proceedings of the 35th International Conference on Machine Learning , pages=. 2018 , organization=

    2018

  34. [42]

    Advances in Neural Information Processing Systems , volume=

    Improving model calibration with accuracy versus uncertainty optimization , author=. Advances in Neural Information Processing Systems , volume=

  35. [43]

    Transactions on Machine Learning Research , pages=

    Meta-Calibration: Learning of Model Calibration Using Differentiable Expected Calibration Error , author=. Transactions on Machine Learning Research , pages=

  36. [44]

    The Ninth International Conference on Learning Representations (ICLR 2021) , year=

    Evaluation of neural architectures trained with square loss vs cross-entropy in classification tasks , author=. The Ninth International Conference on Learning Representations (ICLR 2021) , year=

    2021

  37. [45]

    International Conference on Machine Learning , pages=

    Dual focal loss for calibration , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  38. [46]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Local temperature scaling for probability calibration , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  39. [47]

    Advances in Neural Information Processing Systems , volume=

    Adafocal: Calibration-aware adaptive focal loss , author=. Advances in Neural Information Processing Systems , volume=

  40. [48]

    Nature Medicine , volume=

    Deep learning-aided decision support for diagnosis of skin disease across skin tones , author=. Nature Medicine , volume=. 2024 , publisher=

    2024

  41. [49]

    The Lancet Digital Health , volume=

    Combining the strengths of radiologists and AI for breast cancer screening: a retrospective analysis , author=. The Lancet Digital Health , volume=. 2022 , publisher=

    2022

  42. [50]

    arXiv preprint arXiv:1409.1556 , year=

    Very deep convolutional networks for large-scale image recognition , author=. arXiv preprint arXiv:1409.1556 , year=

    arXiv

  43. [51]

    Journal of Applied Meteorology and Climatology , volume=

    A new vector partition of the probability score , author=. Journal of Applied Meteorology and Climatology , volume=

  44. [52]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  45. [53]

    2009 IEEE conference on computer vision and pattern recognition , pages=

    Imagenet: A large-scale hierarchical image database , author=. 2009 IEEE conference on computer vision and pattern recognition , pages=. 2009 , organization=

    2009

  46. [54]

    International Conference on Machine Learning , year=

    A Novel Characterization of the Population Area Under the Risk Coverage Curve (AURC) and Rates of Finite Sample Estimators , author=. International Conference on Machine Learning , year=. 2410.15361 , archivePrefix=

  47. [55]

    arXiv preprint arXiv:2107.06217 , year=

    What classifiers know what they don't? , author=. arXiv preprint arXiv:2107.06217 , year=

  48. [56]

    Statistics , volume=

    Kernel density estimation for heavy-tailed distributions using the Champernowne transformation , author=. Statistics , volume=. 2005 , publisher=

    2005

  49. [57]

    International Conference on Artificial Intelligence and Statistics , pages=

    Mitigating bias in calibration error estimation , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2022 , organization=

    2022

  50. [58]

    Master's thesis, University of Tront , year=

    Learning Multiple Layers of Features from Tiny Images , author=. Master's thesis, University of Tront , year=

  51. [59]

    Statistics and Computing , year=

    Local Regression and Likelihood , author=. Statistics and Computing , year=

  52. [60]

    1994 , publisher=

    Kernel smoothing , author=. 1994 , publisher=

    1994

  53. [61]

    1987 , publisher=

    Density estimation for statistics and data analysis , author=. 1987 , publisher=

    1987

  54. [62]

    5th International Conference on Learning Representations , year=

    Regularizing neural networks by penalizing confident output distributions , author=. 5th International Conference on Learning Representations , year=

  55. [63]

    IEEE Transactions on information theory , volume=

    On optimum recognition error and reject tradeoff , author=. IEEE Transactions on information theory , volume=. 1970 , publisher=

    1970

  56. [64]

    1965 , publisher=

    Special functions and their applications , author=. 1965 , publisher=

    1965

  57. [65]

    Journal of the American statistical Association , volume=

    Strictly proper scoring rules, prediction, and estimation , author=. Journal of the American statistical Association , volume=. 2007 , publisher=

    2007

  58. [66]

    2018 , publisher=

    Density estimation for statistics and data analysis , author=. 2018 , publisher=

    2018

  59. [67]

    1970 , journal=

    Convex analysis , author=. 1970 , journal=

    1970

  60. [68]

    arXiv preprint arXiv:2402.08384 , year=

    Selective Learning: Towards Robust Calibration with Dynamic Regularization , author=. arXiv preprint arXiv:2402.08384 , year=

  61. [69]

    International Conference on Learning Representations , year=

    Bias-reduced uncertainty estimation for deep neural classifiers , author=. International Conference on Learning Representations , year=

  62. [70]

    Journal of Machine Learning Research , volume=

    Optimal strategies for reject option classifiers , author=. Journal of Machine Learning Research , volume=

  63. [71]

    A Call to Reflect on Evaluation Practices for Failure Detection in Image Classification , booktitle =

    Paul F Jaeger and Carsten Tim L. A Call to Reflect on Evaluation Practices for Failure Detection in Image Classification , booktitle =

  64. [72]

    , author=

    On the Foundations of Noise-free Selective Classification. , author=. Journal of Machine Learning Research , volume=

  65. [73]

    International Conference on Learning Representations , year=

    A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks , author=. International Conference on Learning Representations , year=

  66. [74]

    International conference on machine learning , pages=

    Mitigating neural network overconfidence with logit normalization , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  67. [75]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Understanding the feature norm for out-of-distribution detection , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  68. [76]

    International Conference on Machine Learning , pages=

    Scaling Out-of-Distribution Detection for Real-World Settings , author=. International Conference on Machine Learning , pages=. 2022 , organization=

    2022

  69. [77]

    International Conference on Machine Learning (ICML) , pages =

    Training data-efficient image transformers & distillation through attention , author =. International Conference on Machine Learning (ICML) , pages =. 2021 , organization =

    2021

  70. [78]

    Advances in neural information processing systems , volume=

    Energy-based out-of-distribution detection , author=. Advances in neural information processing systems , volume=

  71. [79]

    International conference on machine learning , pages=

    Mix-n-match: Ensemble and compositional methods for uncertainty calibration in deep learning , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  72. [80]

    Justifying recommendations using distantly-labeled reviews and fine-grained aspects , author=. Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) , pages=

    2019

  73. [81]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Revisiting confidence estimation: Towards reliable failure prediction , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  74. [82]

    The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=

    Overcoming Common Flaws in the Evaluation of Selective Classification Systems , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=

  75. [83]

    international conference on machine learning , pages=

    Confidence-aware learning for deep neural networks , author=. international conference on machine learning , pages=. 2020 , organization=

    2020

  76. [84]

    Sociological Methods & Research , volume=

    Fairness in criminal justice risk assessments: The state of the art , author=. Sociological Methods & Research , volume=. 2021 , publisher=

    2021

  77. [85]

    Advances in neural information processing systems , volume=

    Calibration tests in multi-class classification: A unifying framework , author=. Advances in neural information processing systems , volume=

  78. [86]

    Journal of Machine Learning Research , volume=

    Metrics of calibration for probabilistic predictions , author=. Journal of Machine Learning Research , volume=

  79. [87]

    IEEE Transactions on computers , volume=

    On the choice of smoothing parameters for Parzen estimators of probability density functions , author=. IEEE Transactions on computers , volume=. 1976 , publisher=

    1976

  80. [88]

    , author=

    Measuring calibration in deep learning. , author=. CVPR workshops , volume=

  81. [89]

    The Twelfth International Conference on Learning Representations , year=

    Smooth ECE: Principled Reliability Diagrams via Kernel Smoothing , author=. The Twelfth International Conference on Learning Representations , year=

  82. [90]

    Conference on Learning Theory , pages=

    The geometry of losses , author=. Conference on Learning Theory , pages=. 2014 , organization=

    2014

  83. [91]

    International Conference on Artificial Intelligence and Statistics , pages=

    Accuracy-Preserving Calibration via Statistical Modeling on Probability Simplex , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2024 , organization=

    2024

  84. [92]

    Fifteenth International Conference on Machine Vision (ICMV 2022) , volume=

    Confidence-aware calibration and scoring functions for curriculum learning , author=. Fifteenth International Conference on Machine Vision (ICMV 2022) , volume=. 2023 , organization=

    2022

  85. [93]

    Proceedings of the IEEE international conference on computer vision , pages=

    Focal loss for dense object detection , author=. Proceedings of the IEEE international conference on computer vision , pages=

  86. [94]

    Monthly weather review , volume=

    Verification of forecasts expressed in terms of probability , author=. Monthly weather review , volume=

  87. [95]

    Advances in neural information processing systems , volume=

    Selective classification for deep neural networks , author=. Advances in neural information processing systems , volume=

  88. [96]

    The Eleventh International Conference on Learning Representations , year=

    What Can we Learn From The Selective Prediction And Uncertainty Estimation Performance Of 523 Imagenet Classifiers? , author=. The Eleventh International Conference on Learning Representations , year=

  89. [97]

    International Conference on artificial intelligence and statistics , pages=

    Active learning under label shift , author=. International Conference on artificial intelligence and statistics , pages=. 2021 , organization=

    2021

  90. [98]

    arXiv preprint arXiv:2404.13016 , year=

    Optimizing calibration by gaining aware of prediction correctness , author=. arXiv preprint arXiv:2404.13016 , year=

  91. [99]

    International Conference on Machine Learning , pages=

    Fast differentiable sorting and ranking , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  92. [100]

    Proceedings of the 30th ACM International Conference on Multimedia , pages=

    Angular gap: Reducing the uncertainty of image difficulty through model calibration , author=. Proceedings of the 30th ACM International Conference on Multimedia , pages=

  93. [101]

    Proceedings of the 26th annual international conference on machine learning , pages=

    Curriculum learning , author=. Proceedings of the 26th annual international conference on machine learning , pages=

  94. [102]

    Advances in large margin classifiers , volume=

    Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods , author=. Advances in large margin classifiers , volume=. 1999 , publisher=

    1999

  95. [103]

    Advances in neural information processing systems , volume=

    Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with dirichlet calibration , author=. Advances in neural information processing systems , volume=

  96. [104]

    Artificial intelligence and statistics , pages=

    Beta calibration: a well-founded and easily implemented improvement on logistic calibration for binary classifiers , author=. Artificial intelligence and statistics , pages=. 2017 , organization=

    2017

  97. [105]

    International Conference on Artificial Intelligence and Statistics , pages=

    Non-parametric calibration for classification , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2020 , organization=

    2020

  98. [106]

    International Conference on Artificial Intelligence and Statistics , pages=

    Consistent and Asymptotically Unbiased Estimation of Proper Calibration Errors , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2024 , organization=

    2024

  99. [107]

    Information fusion , volume=

    A review of uncertainty quantification in deep learning: Techniques, applications and challenges , author=. Information fusion , volume=. 2021 , publisher=

    2021

  100. [108]

    Proceedings of NAACL-HLT , pages=

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , author=. Proceedings of NAACL-HLT , pages=

  101. [109]

    Johns Hopkins University, Tech

    Sampling uniformly from the unit simplex , author=. Johns Hopkins University, Tech. Rep , volume=

  102. [110]

    CoRR , year=

    Rethinking Early Stopping: Refine, Then Calibrate , author=. CoRR , year=

  103. [111]

    Advances in Neural Information Processing Systems , volume=

    Rethinking calibration of deep neural networks: Do not be afraid of overconfidence , author=. Advances in Neural Information Processing Systems , volume=

  104. [112]

    ECAI 2024 , pages=

    Improving Calibration by Relating Focal Loss, Temperature Scaling, and Properness , author=. ECAI 2024 , pages=. 2024 , publisher=

    2024

  105. [113]

    Journal of statistical planning and inference , volume=

    Improving predictive inference under covariate shift by weighting the log-likelihood function , author=. Journal of statistical planning and inference , volume=. 2000 , publisher=

    2000

  106. [114]

    China National Conference on Chinese Computational Linguistics , pages=

    A robustly optimized BERT pre-training approach with post-training , author=. China National Conference on Chinese Computational Linguistics , pages=. 2021 , organization=

    2021

  107. [115]

    , author=

    DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. , author=. Proceedings of Thirty-third Conference on Neural Information Processing Systems (NIPS2019) , year=

Pith tools

Reviewed June 30, 2026 · model on record in the stance chip above.