Pith. sign in

REVIEW 3 major objections 5 minor 67 references

Risk-aware Classification via Uncertainty Quantification

T0 review · 3 major / 5 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 claims that augmenting Evidential Deep Learning with per-sample pignistic priors reduces misclassification cost by at least 55% on MNIST and 38% on CIFAR10 compared with cost-sensitive training, while keeping test accuracy…

desk verdict A modest but useful extension of EDL, where the fine-tuning and fusion evidence in Section 3 is the real new contribution; the risk-aware variants are largely a reprise of WACV 2021, and the headline cost-reduction numbers need seeds and a corrected bandit story. read the letter →

arxiv 2412.03391 v1 pith:FDYPAUPH submitted 2024-12-04 cs.LG

Murat Sensoy , Lance M. Kaplan , Simon Julier , Maryam Saleki , Federico Cerutti This is my paper · ORCID
classification cs.LG MSC 68T0168T45 PACS 07.05.Mh
keywords risk-awareclassificationevidentialdeeplearninguncertaintyquantificationDirichletdistributionpignisticprobabilitiescost-sensitivepolicygradientcompositionality
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 sets out to make deep classifiers risk-aware: instead of only maximizing accuracy, the classifier should minimize the expected cost of its mistakes, especially when some errors are far more expensive than others. The authors argue that Evidential Deep Learning (EDL), which replaces softmax probabilities with a Dirichlet distribution over class probabilities, is the right base because it already satisfies three desiderata: it quantifies epistemic uncertainty, it can fine-tune pretrained models, and it can fuse classifiers trained on disjoint label sets. They then extend EDL with a per-sample pignistic prior, a learned redistribution of prior evidence, so that the model's decisions are biased toward low-risk categories mainly when the evidence is weak. On MNIST and CIFAR10, the extended methods reduce misclassification cost by at least 55% and 38% respectively relative to cost-sensitive training, with accuracy essentially unchanged. A policy-gradient variant, EDL(pg), is presented as learning from bandit feedback, although the stated update assumes access to the true label.

What carries the argument

The load-bearing object is the Dirichlet distribution as a conjugate prior over class probabilities, parameterised by $\alpha = c_\theta(x)+1$ where $c_\theta(x)$ is a non-negative evidence vector from the network. Risk-awareness is injected through a second head that produces a per-sample pignistic prior $\gamma_{\Theta}(x)=K\cdot\mathrm{softmax}(W g_\theta(x)+b)$ with total mass $K$, so that the expected risk formula in Eq. (9) weighs each class by its risk entry $R_{y i}$ times the sum of evidence and prior counts, normalised by $K+\sum_j c_{\theta j}(x)$. This construction is what lets the model hedge: when evidence is abundant the prior barely matters, and when evidence is scarce the prior steers decisions toward low-cost categories. A policy-gradient update derived from Eq. (12) is the mechanism claimed to enable bandit-feedback training.

What would settle it

Run EDL(pg) in a true bandit environment where only the cost of the chosen class is revealed and the true label is never shown; if the learner cannot compute the gradient in Algorithm 1 (which multiplies R_{y i} by the score-function term), the claimed bandit learning collapses. A simpler check: inspect the update at line 4 of Algorithm 1 and verify that R_{y i} requires the true y to be known; the paper's Section 4.3 admits 'the (x, y) tuple represents each sample x and its true label,' so a reader can settle the matter by seeing whether the observed cost alone determines R_{y i} for every i.

Watch

Extended reading notes

Core claim

The central claim is that EDL's Dirichlet output can be turned into a risk-aware decision rule by learning a per-sample pignistic prior $\gamma_{\Theta}(x)=K\cdot\mathrm{softmax}(W g_\theta(x)+b)$ that is added to the evidence counts before computing expected risk. The expected risk of labelling $x$ as the true class $y$ is $E[\mathrm{risk}(x)] = \frac{\sum_i R_{y i}(c_{\theta i}(x)+\gamma_{\Theta i}(x))}{K+\sum_j c_{\theta j}(x)}$, and the network can be trained to minimise this quantity along with EDL's loss. Three variants are proposed: riskEDL trains both heads jointly, EDL(p) freezes the trained EDL network and fits only the pignistic prior, and EDL(pg) applies a policy-gradient (REINFORCE-style) update to the prior. The paper reports that EDL(p) and EDL(pg) reduce misclassification cost by at least 55% on MNIST and 38% on CIFAR10 compared to cost-sensitive cross-entropy training (CS-L2), while keeping test accuracy essentially unchanged (Tables 2 and 3). It also claims that fine-tuning pretrained classifiers with EDL loss improves their uncertainty estimates and enables principled fusion of classifiers on disjoint label sets.

Load-bearing premise

The bandit-feedback claim for EDL(pg) assumes the learner can observe the true class label y to compute the risk R_{y i} of the chosen action, even though the stated setting only reveals the cost of the action actually taken.

Editorial extensions

If this is right

  • Fine-tuning a pretrained classifier with EDL loss preserves its accuracy while making its uncertainty much higher on misclassified and out-of-distribution inputs; fusing two EDL-tuned classifiers on disjoint label sets raises combined accuracy from 62.4% to 73.5% for CIFAR10+CIFAR100.
  • Risk-aware EDL with pignistic priors reduces misclassification cost by at least 55% on MNIST and 38% on CIFAR10 relative to cost-sensitive training, with accuracy essentially unchanged.
  • EDL(p) and EDL(pg) outperform riskEDL on the tested risk matrices because they apply the risk correction mostly when evidence is low; riskEDL may be preferable when the risk matrix is binary with equal error costs.
  • EDL(pg) reaches lower cost than cost-sensitive training within a few training epochs and approaches the performance of EDL(p), which sees the full cost matrix.

Reading between the lines

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

  • If the bandit claim fails the label-access test, the practical scope of EDL(pg) shrinks to settings where the true label is recoverable from the observed cost; a genuine bandit variant would need to estimate the cost of unchosen actions or use importance-weighted updates.
  • The same pignistic-prior construction could be grafted onto other Dirichlet-predicting uncertainty models (e.g., prior networks), since the risk formula only needs the Dirichlet parameters; testing that would separate the benefit of the prior mechanism from EDL-specific loss details.
  • The method's apparent ability to shift errors toward low-cost confusions while preserving accuracy suggests it could be directly useful for ordinal or auction-like tasks, where overestimating value is much costlier than underestimating it.
  • The three desiderata (D1-D3) are presented as necessary but not sufficient; methods that require architectural overhauls or external outlier datasets will struggle to satisfy D2 and D3, which may be a useful filter for comparing uncertainty-quantification approaches.
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 three desiderata for real-world risk-aware classification (D1: Dirichlet outputs with epistemic uncertainty, D2: transfer from pre-trained models, D3: compositionality), argues that Evidential Deep Learning satisfies them, and presents experiments showing that EDL-tuned pre-trained models improve entropy-based uncertainty scores and that two EDL classifiers can be fused. The main extension is a per-sample pignistic prior gamma_Theta(x) = K softmax(W g_theta(x) + b), with three training variants: riskEDL, EDL(p), and EDL(pg) (a REINFORCE/bandit-style update). On MNIST and CIFAR10 with LeNet, the risk-aware variants are reported to reduce misclassification cost by roughly 55% and 38% relative to cost-sensitive cross-entropy while maintaining accuracy.

Significance. If the empirical claims hold, this is a practically useful and conceptually clean extension of EDL: the pignistic-prior mechanism is simple, Eq. (9) is correctly derived from Dirichlet means, and the paper includes an external cost-sensitive baseline as well as concrete D2/D3 demonstrations. The policy-gradient variant targets a realistic feedback setting, which strengthens the contribution. However, the headline cost reductions rest on single-run point estimates on two small datasets, and the bandit variant as written appears to require true labels; these issues must be resolved before the central claims can be accepted.

major comments (3)
  1. [Section 4, Eq. (9)] The quantity in Eq. (9) is the expected cost of a randomized decision rule that draws the predicted label i with probability p_i. This is not the expected cost of a deterministic decision j, which is sum_i R_{i,j} p_i and is minimized by argmin_j. The paper never states the decision rule used to produce the test-set costs in Tables 2 and 3; Section 5 says the expectation of q_theta(p|x) is used 'as the predictive distribution,' which suggests argmax of the mean pignistic probabilities. If so, the training objective in Eq. (9) does not match the evaluation metric; if instead labels are sampled from P(i|x), the reported costs are random and need variance estimates. Please specify the test-time rule and, if needed, optimize the objective corresponding to the actual decision rule.
  2. [Section 4.3, Algorithm 1] Algorithm 1 is presented as learning from bandit feedback, but it iterates over (x,y) pairs and uses R_{y,i} in the update. In a true bandit the learner observes only the scalar cost c_i = R_{y,i} of the sampled action i; a REINFORCE update with c_i * grad log P(i|x) is unbiased and does not require y. The pseudocode as written requires the true label y (or the full risk row), so the implementation may be supervised, and the paper provides no ablation showing that EDL(pg) works when only the chosen action's cost is available. Please clarify the feedback model and evaluate the bandit claim under genuinely scalar feedback.
  3. [Section 5, Tables 2 and 3] The central empirical claim—that the proposed approaches reduce misclassification cost by at least 55% on MNIST and 38% on CIFAR10 relative to CS-L2—is based on one run per method with no standard deviations, no seeds, and no released code. With differences of a few tenths of a percent in accuracy and cost reductions that could be influenced by initialization or the stochastic policy in EDL(pg), the reported point estimates cannot support a general superiority claim. Please provide multi-seed means and deviations (and ideally paired significance tests), and specify how the CS-L2 baseline was tuned so the comparison is fair.
minor comments (5)
  1. [Section 2.3] The desiderata are introduced as 'necessary' without an independent argument; since the paper sets them up so that EDL satisfies them by construction, consider framing them as design goals rather than foundational requirements.
  2. [Section 5] In the MNIST risk-matrix definition, 'misclassifing' should be 'misclassifying', and the prose should clarify whether 'overestimating' means predicting a digit with a larger value than the true digit.
  3. [Section 5, Tables 2 and 3] The label 'L2' for standard cross-entropy training is confusing because L2 usually denotes squared error; use 'CE' or 'cross-entropy' instead.
  4. [Section 4.3] Equation (11) writes the update for theta, while Algorithm 1 updates Theta = {W, b}; clarify that only the pignistic-prior parameters are trained for EDL(pg) and the evidence network is frozen.
  5. [Section 3.2, Table 1] The out-of-distribution source for the CIFAR100 row is not stated; please specify which dataset was used as OoD for that row.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity: the cost-reduction claim is independently benchmarked; only D1 is satisfied by definition.

  1. self definitional [Section 2.3 (Desiderata) vs. Section 3 (EDL definition)]
    "“D1 : the output of a classification system should be interpreted as a Dirichlet distribution with some estimation of its epistemic uncertainty” (Sec. 2.3); “Such results can be the base for the parameters α_i = c_θ(x_i) + 1 of a Dirichlet distribution” (Sec. 3)."

    D1 is defined as 'output interpreted as a Dirichlet distribution,' and EDL is defined as producing Dirichlet parameters by adding 1 to the network's non-negative evidence vector. The paper then presents the satisfaction of D1 as a demonstrated unity between foundational desiderata and EDL's operational attributes. For D1 this is true by construction, not by derivation: the desideratum is a restatement of EDL's architectural definition. D2 and D3 have empirical content, and the headline cost-reduction claim in Tables 2-3 is an external comparison against L2, CS-L2, and EDL under fixed risk matrices, so this tautology is framing rather than the engine of the central quantitative result.

full rationale

The paper's strongest quantitative claim is the misclassification-cost reduction in Section 5 (Tables 2 and 3), where riskEDL, EDL(p), and EDL(pg) are compared with standard cross-entropy, cost-sensitive cross-entropy (CS-L2), and plain EDL under two fixed risk matrices. This comparison is externally benchmarked and statistically falsifiable; the reported cost is the test-set value of the same risk objective, but that is normal supervised cost-sensitive learning, not a fitted input renamed as a prediction. The Dirichlet expectation in Eq. (9) is standard conjugate-prior mathematics, and EDL(pg) is explicitly an adaptation of REINFORCE. The main circularity-adjacent element is the framing of D1-D3: D1 is defined as 'output interpreted as a Dirichlet distribution,' which is exactly what EDL outputs by construction, so the paper's 'demonstration' that EDL fulfills D1 is a restatement rather than a derivation. D2 and D3 are supported by experiments (Table 1, Figure 4) rather than derived. The self-citations to [SKK18a] and [SSJ+21] provide the base EDL loss and preliminary riskEDL result, but the central cost-reduction claim is not obtained by citing them; it rests on the new empirical comparison. The EDL(pg) bandit-feedback claim is weakened by Algorithm 1's use of the true label y via R_yi, and the lack of seeds/released code is a reproducibility concern, but neither is a circularity of the derivation.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The ledger shows the paper's main new component is the per-sample pignistic prior, which is a learned function rather than a first-principles object. The desiderata are self-defined. The Dirichlet machinery and EDL loss are carried from prior work, and the risk-aware training uses a hand-set weight kappa.

free parameters (3)
  • kappa = 0.01
    Weight of the risk-awareness loss in riskEDL (Section 4.1). Set by hand for all experiments; affects the trade-off between EDL loss and risk loss.
  • clamping_threshold = 10
    Upper bound in evidence function exp(min(x,10)) + (x - bg(x)) used for fine-tuning pretrained models (Section 3.2). Chosen for numerical stability.
  • annealing_coefficient = min(1.0, t/10)
    Annealing schedule in the EDL KL term (Equation 4), carried from [SKK18a]. Hand-specified to delay the effect of the KL regularization.
assumptions (5)
  • standard math Dirichlet-multinomial conjugacy: posterior of a multinomial with Dirichlet prior is Dirichlet with summed parameters.
    Used in Section 2.2 and throughout to justify EDL's posterior interpretation.
  • domain assumption Pignistic probabilities p are modeled as Dirichlet distributed with parameters c_theta(x) + gamma_Theta(x) (Equation 7).
    Central modeling choice for risk-aware EDL; not derived from data or first principles.
  • domain assumption Expected misclassification risk is the expectation of the cost under the Dirichlet mean of p (Equation 9).
    Assumes cost is linear in pignistic probabilities and that the Dirichlet expectation is the right risk measure.
  • ad hoc to paper Desiderata D1-D3 are necessary for real-world risk-aware classification (Section 2.3).
    Stated as foundational without proof; the paper even notes they are not sufficient. They are designed so EDL satisfies them.
  • domain assumption Evidence from classifiers trained on disjoint categories can be fused by concatenating Dirichlet parameters (Section 3.3).
    Relies on aggregation/neutrality of Dirichlet distributions, but also requires commensurable evidence scales; empirically holds after EDL tuning but not for the original pretrained classifiers.
invented entities (1)
  • Per-sample pignistic prior gamma_Theta(x) = K softmax(W g_theta(x) + b)
    purpose: Redistributes the uniform Dirichlet prior counts based on the input sample so that, when evidence is weak, the model favors lower-risk categories.
    Introduced in Equation 8 and trained on the paper's datasets; no external validation outside the reported MNIST/CIFAR10 experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Risk-aware Classification via Uncertainty Quantification." pith.science (2026). https://pith.science/paper/FDYPAUPH

@misc{pith2026241203391,
  author       = {Pith},
  title        = {Pith review of: Risk-aware Classification via Uncertainty Quantification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FDYPAUPH}},
  note         = {Machine review of arXiv:2412.03391}
}
read the original abstract

Autonomous and semi-autonomous systems are using deep learning models to improve decision-making. However, deep classifiers can be overly confident in their incorrect predictions, a major issue especially in safety-critical domains. The present study introduces three foundational desiderata for developing real-world risk-aware classification systems. Expanding upon the previously proposed Evidential Deep Learning (EDL), we demonstrate the unity between these principles and EDL's operational attributes. We then augment EDL empowering autonomous agents to exercise discretion during structured decision-making when uncertainty and risks are inherent. We rigorously examine empirical scenarios to substantiate these theoretical innovations. In contrast to existing risk-aware classifiers, our proposed methodologies consistently exhibit superior performance, underscoring their transformative potential in risk-conscious classification strategies.

Figures

Figures reproduced from arXiv: 2412.03391 by the authors.

Figure 1
Figure 1. Classifying a rotated digit 1 (positioned at the bottom) across various angles spanning from 0 to 180 degrees involves the calculation of classification probabilities through the softmax function. The maximization of the multinomial likelihood with respect to the neural network parameters θ is accomplished with a preference for the computational convenience offered by the equivalent problem of minimizing the negativ… view at source ↗
Figure 2
Figure 2. Probability density function of different [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Visual depiction of the plate notation for discriminative evidential classifiers. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: ROC (left) and PRC (right) for the fusion of pre-trained and [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Proposed architecture for risk-aware classification. [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Risk matrices. standard LeNet with ReLU. All experiments are implemented using Tensor￾flow [ABC+16], and training utilizes the Adam optimizer [KB15] with default settings. For the MNIST dataset, we train the LeNet architecture [LBB+98] with 20 and 50 filters of size 5 …
Figure 6
Figure 6. Figure 6: For the MNIST dataset, the risk matrix is such that [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 7
Figure 7. Figure 7: Change in the misclassification cost for MNIST (a) and CIFAR10 (b) test sets [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 62 canonical work pages

  1. [1]

    Abadi, P

    M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, M. Isard, et al. Tensorflow: A system for large-scale machine learning. In OSDI , volume 16, pages 265--283, 2016

    work page 2016

  2. [2]

    Barf: A new direct and cross-based binary residual feature fusion with uncertainty-aware module for medical image classification

    Moloud Abdar, Mohammad Amin Fahami, Satarupa Chakrabarti, Abbas Khosravi, Pawe P awiak, U Rajendra Acharya, Ryszard Tadeusiewicz, and Saeid Nahavandi. Barf: A new direct and cross-based binary residual feature fusion with uncertainty-aware module for medical image classification. Information Sciences , 577:353--378, 2021

    work page 2021

  3. [3]

    Rajendra Acharya, Vladimir Makarenkov, and Saeid Nahavandi

    Moloud Abdar, Farhad Pourpanah, Sadiq Hussain, Dana Rezazadegan, Li Liu, Mohammad Ghavamzadeh, Paul Fieguth, Xiaochun Cao, Abbas Khosravi, U. Rajendra Acharya, Vladimir Makarenkov, and Saeid Nahavandi. A review of uncertainty quantification in deep learning: Techniques , applications and challenges. Information Fusion , 76:243--297, December 2021

    work page 2021

  4. [4]

    Uncertaintyfusenet: robust uncertainty-aware hierarchical feature fusion model with ensemble monte carlo dropout for covid-19 detection

    Moloud Abdar, Soorena Salari, Sina Qahremani, Hak-Keung Lam, Fakhri Karray, Sadiq Hussain, Abbas Khosravi, U Rajendra Acharya, Vladimir Makarenkov, and Saeid Nahavandi. Uncertaintyfusenet: robust uncertainty-aware hierarchical feature fusion model with ensemble monte carlo dropout for covid-19 detection. Information Fusion , 90:364--381, 2023

    work page 2023

  5. [5]

    Deep evidential regression

    Alexander Amini, Wilko Schwarting, Ava Soleimany, and Daniela Rus. Deep evidential regression. In H. Larochelle, M. Ranzato, R. Hadsell, M. F. Balcan, and H. Lin, editors, Advances in Neural Information Processing Systems , volume 33, pages 14927--14937, 2020

    work page 2020

  6. [6]

    Blundell, J

    C. Blundell, J. Cornebise, K. Kavukcuoglu, and D. Wiestra. Weight uncertainty in neural networks. In ICML , 2015

    work page 2015

  7. [7]

    Deep, spatially coherent inverse sensor models with uncertainty incorporation using the evidential framework

    Daniel Bauer, Lars Kuhnert, and Lutz Eckstein. Deep, spatially coherent inverse sensor models with uncertainty incorporation using the evidential framework. In 2019 IEEE Intelligent Vehicles Symposium (IV) , pages 2490--2495. IEEE, 2019

    work page 2019

  8. [8]

    Evidential deep learning for open set action recognition

    Wentao Bao, Qi Yu, and Yu Kong. Evidential deep learning for open set action recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 13349--13358, 2021

    work page 2021

Show all 67 references
  1. [9]

    Kaplan, and Murat S ensoy

    Federico Cerutti, Lance M. Kaplan, and Murat S ensoy. Evidential Reasoning and Learning : A Survey . In Thirty- First International Joint Conference on Artificial Intelligence , volume 6, pages 5418--5425, July 2022

    2022

  2. [10]

    Cost-aware pre-training for multiclass cost-sensitive deep learning

    Yu-An Chung, Hsuan-Tien Lin, and Shao-Wen Yang. Cost-aware pre-training for multiclass cost-sensitive deep learning. In Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16) , pages 1411--1417, 2016

    2016

  3. [11]

    Dempster

    A.P. Dempster. A generalization of Bayesian inference. In Classic Works of the Dempster-Shafer Theory of Belief Functions , pages 73--104. Springer , 2008

    2008

  4. [12]

    A definition of subjective possibility

    Didier Dubois, Henri Prade, and Philippe Smets. A definition of subjective possibility. International Journal of Approximate Reasoning , 48(2):352--364, 2008

    2008

  5. [13]

    The foundations of cost-sensitive learning

    Charles Elkan. The foundations of cost-sensitive learning. In In Proceedings of the Seventeenth International Joint Conference on Artificial Intelligence , pages 973--978, 2001

    2001

  6. [14]

    Introduction to the dirichlet distribution and related processes

    Bela A Frigyik, Amol Kapila, and Maya R Gupta. Introduction to the dirichlet distribution and related processes. Department of Electrical Engineering, University of Washignton, UWEETR-2010-0006 , (0006):1--27, 2010

    2010

  7. [15]

    Exploring the limits of out-of-distribution detection

    Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution detection. Advances in Neural Information Processing Systems , 34:7068--7081, 2021

    2021

  8. [16]

    Dense out-of-distribution detection by robust learning on synthetic negative data

    Matej Grci \'c , Petra Bevandi \'c , Zoran Kalafati \'c , and Sini s a S egvi \'c . Dense out-of-distribution detection by robust learning on synthetic negative data. Sensors , 24(4), 2024

    2024

  9. [17]

    Cost-sensitive regularization for diabetic retinopathy grading from eye fundus images

    Adrian Galdran, Jose Dolz, Hadi Chakor, Herv \'e Lombaert, and Ismail Ben Ayed. Cost-sensitive regularization for diabetic retinopathy grading from eye fundus images. In International Conference on Medical Image Computing and Computer-Assisted Intervention , pages 665--674. Sp...

    2020

  10. [18]

    Gal and Z

    Y. Gal and Z. Ghahramani. Dropout as a B ayesian approximation: Representing model uncertainty in deep learning. In ICML , 2016

    2016

  11. [19]

    Quantifying and leveraging classification uncertainty for chest radiograph assessment

    Florin C Ghesu, Bogdan Georgescu, Eli Gibson, Sebastian Guendel, Mannudeep K Kalra, Ramandeep Singh, Subba R Digumarthy, Sasa Grbic, and Dorin Comaniciu. Quantifying and leveraging classification uncertainty for chest radiograph assessment. In International Conference on Medic...

    2019

  12. [20]

    On calibration of modern neural networks

    Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q Weinberger. On calibration of modern neural networks. In International Conference on Machine Learning , pages 1321--1330. PMLR, 2017

    2017

  13. [21]

    Kaplan, Audun Josang, Dong H

    Zhen Guo, Zelin Wan, Qisheng Zhang, Xujiang Zhao, Qi Zhang, Lance M. Kaplan, Audun Josang, Dong H. Jeong, Feng Chen, and Jin-Hee Cho. A survey on uncertainty reasoning and quantification in belief theory and its application to deep learning. Information Fusion , 101, 2024

    2024

  14. [22]

    Uncertainty-aware reliable text classification

    Yibo Hu and Latifur Khan. Uncertainty-aware reliable text classification. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages 628--636, 2021

    2021

  15. [23]

    u hl, and Jakob Sch \

    Patrick Hemmer, Niklas K \"u hl, and Jakob Sch \"o ffer. Deal: deep evidential active learning for image classification. In Deep Learning Applications, Volume 3 , pages 171--192. Springer, 2022

    2022

  16. [24]

    Aleatory and epistemic uncertainty in probability elicitation with an example from hazardous waste management

    Stephen C Hora. Aleatory and epistemic uncertainty in probability elicitation with an example from hazardous waste management. Reliability Engineering & System Safety , 54(2):217--223, 1996

    1996

  17. [25]

    Noise contrastive priors for functional uncertainty

    Danijar Hafner, Dustin Tran, Alex Irpan, Timothy Lillicrap, and James Davidson. Noise contrastive priors for functional uncertainty. 2019

    2019

  18. [26]

    Aleatoric and Epistemic Uncertainty in Machine Learning : An Introduction to Concepts and Methods

    Eyke H \"u llermeier and Willem Waegeman. Aleatoric and Epistemic Uncertainty in Machine Learning : An Introduction to Concepts and Methods . Machine Learning , 110(3):457--506, March 2021

    2021

  19. [27]

    Learning from bandit feedback: An overview of the state-of-the-art

    Olivier Jeunen, Dmytro Mykhaylov, David Rohde, Flavian Vasile, Alexandre Gilotte, and Martin Bompaire. Learning from bandit feedback: An overview of the state-of-the-art. In Proceedings of the ACM RecSys Workshop on Reinforcement Learning and Robust Estimators for Recommendati...

    2019

  20. [28]

    A. Josang. Subjective Logic: A Formalism for Reasoning Under Uncertainty . Springer, 2016

    2016

  21. [29]

    Kingma and J

    D.P. Kingma and J. Ba. Adam: A method for stochastic optimisation. In ICLR , 2015

    2015

  22. [30]

    S. Kotz, N. Balakrishnan, and N.L. Johnson. Continuous Multivariate Distributions , volume 1. Wiley, New York, 2000

    2000

  23. [31]

    Kendall and Y

    A. Kendall and Y. Gal. What uncertainties do we need in B ayesian deep learning for computer vision? In NIPS , 2017

    2017

  24. [32]

    What uncertainties do we need in bayesian deep learning for computer vision? In Advances in neural information processing systems , pages 5574--5584, 2017

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? In Advances in neural information processing systems , pages 5574--5584, 2017

    2017

  25. [33]

    Cost-sensitive learning of deep feature representations from imbalanced data

    Salman H Khan, Munawar Hayat, Mohammed Bennamoun, Ferdous A Sohel, and Roberto Togneri. Cost-sensitive learning of deep feature representations from imbalanced data. IEEE transactions on neural networks and learning systems , 29(8):3573--3587, 2017

    2017

  26. [34]

    Cost-sensitive learning with neural networks

    Matjaz Kukar, Igor Kononenko, et al. Cost-sensitive learning with neural networks. In ECAI , pages 445--449, 1998

    1998

  27. [35]

    Gradient-based learning applied to document recognition

    Yann LeCun, L \'e on Bottou, Yoshua Bengio, Patrick Haffner, et al. Gradient-based learning applied to document recognition. In the IEEE , 86(11):2278--2324, 1998

    1998

  28. [36]

    LeCun, P

    Y. LeCun, P. Haffner, L. Bottou, and Y. Bengio. Object recognition with gradient-based learning. In Shape, Contour and Grouping in Computer Vision , 1999

    1999

  29. [37]

    Simple and principled uncertainty estimation with deterministic deep learning via distance awareness

    Jeremiah Liu, Zi Lin, Shreyas Padhy, Dustin Tran, Tania Bedrax Weiss, and Balaji Lakshminarayanan. Simple and principled uncertainty estimation with deterministic deep learning via distance awareness. Advances in Neural Information Processing Systems , 33:7498--7512, 2020

    2020

  30. [38]

    Lakshminarayanan, A

    B. Lakshminarayanan, A. Pritzel, and C. Blundell. Simple and scalable predictive uncertainty estimation using deep ensembles. In NIPS , 2017

    2017

  31. [39]

    Reliability analysis for finger movement recognition with raw electromyographic signal by evidential convolutional networks

    Yuzhou Lin, Ramaswamy Palaniappan, Philippe De Wilde, and Ling Li. Reliability analysis for finger movement recognition with raw electromyographic signal by evidential convolutional networks. IEEE Transactions on Neural Systems and Rehabilitation Engineering , 30:96--107, 2022

    2022

  32. [40]

    Louizos and M

    C. Louizos and M. Welling. Multiplicative normalizing flows for variational bayesian neural networks. In ICML , 2017

    2017

  33. [41]

    Predictive uncertainty estimation via prior networks

    Andrey Malinin and Mark Gales. Predictive uncertainty estimation via prior networks. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors, Advances in Neural Information Processing Systems 31 , pages 7047--7058. 2018

    2018

  34. [42]

    Reverse kl-divergence training of prior networks: Improved uncertainty and adversarial robustness

    Andrey Malinin and Mark JF Gales. Reverse kl-divergence training of prior networks: Improved uncertainty and adversarial robustness. 2019

    2019

  35. [43]

    Towards neural networks that provably know when they don't know

    Alexander Meinke and Matthias Hein. Towards neural networks that provably know when they don't know. In ICLR 2020 , 2020

    2020

  36. [44]

    Calibrating deep neural networks using focal loss

    Jishnu Mukhoti, Viveka Kulharia, Amartya Sanyal, Stuart Golodetz, Philip Torr, and Puneet Dokania. Calibrating deep neural networks using focal loss. Advances in Neural Information Processing Systems , 33, 2020

    2020

  37. [45]

    Deterministic neural networks with appropriate inductive biases capture epistemic and aleatoric uncertainty

    Jishnu Mukhoti, Andreas Kirsch, Joost van Amersfoort, Philip HS Torr, and Yarin Gal. Deterministic neural networks with appropriate inductive biases capture epistemic and aleatoric uncertainty. arXiv preprint arXiv:2102.11582 , 2021

    2021 arXiv

  38. [46]

    Machine learning: a probabilistic perspective

    Kevin P Murphy. Machine learning: a probabilistic perspective . MIT press, 2012

    2012

  39. [47]

    Deep deterministic uncertainty for semantic segmentation

    Jishnu Mukhoti, Joost van Amersfoort, Philip HS Torr, and Yarin Gal. Deep deterministic uncertainty for semantic segmentation. arXiv preprint arXiv:2111.00079 , 2021

    2021 arXiv

  40. [48]

    Pe 16-007

    NHTSA. Pe 16-007. technical report, u.s. department of transportation, national highway traffic safety administration, jan 2017. tesla crash preliminary evaluation report. 2016

    2017

  41. [49]

    Epistemic uncertainty quantification in deep learning classification by the delta method

    Geir K Nilsen, Antonella Z Munthe-Kaas, Hans J Skaug, and Morten Brun. Epistemic uncertainty quantification in deep learning classification by the delta method. Neural networks , 145:164--176, 2022

    2022

  42. [50]

    Implicit weight uncertainty in neural networks

    Nick Pawlowski, Andrew Brock, Matthew CH Lee, Martin Rajchl, and Ben Glocker. Implicit weight uncertainty in neural networks. In Bayesian Deep Learning, a NIPS 2017 Workshop , 2017

    2017

  43. [51]

    Rahul Rahaman and Alexandre H. Thiery. Uncertainty quantification and deep ensembles. In A. Beygelzimer, Y. Dauphin, P. Liang, and J. Wortman Vaughan, editors, Advances in Neural Information Processing Systems , 2021

    2021

  44. [52]

    Mcua: Multi-level context and uncertainty aware dynamic deep ensemble for breast cancer histology image classification

    Zakaria Senousy, Mohammed M Abdelsamea, Mohamed Medhat Gaber, Moloud Abdar, U Rajendra Acharya, Abbas Khosravi, and Saeid Nahavandi. Mcua: Multi-level context and uncertainty aware dynamic deep ensemble for breast cancer histology image classification. IEEE Transactions on Bio...

    2021

  45. [53]

    Evidential deep learning for guided molecular property prediction and discovery

    Ava P Soleimany, Alexander Amini, Samuel Goldman, Daniela Rus, Sangeeta N Bhatia, and Connor W Coley. Evidential deep learning for guided molecular property prediction and discovery. ACS central science , 7(8):1356--1367, 2021

    2021

  46. [54]

    Reinforcement learning: An introduction

    Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction . MIT press, 2018

    2018

  47. [55]

    Medspecsearch: Medical specialty search

    Mehmet Ulu c S ahin, Eren Balatkan, Cihan Eran, Engin Zeydan, and Reyyan Yeniterzi. Medspecsearch: Medical specialty search. In European Conference on Information Retrieval , pages 225--229. Springer, 2019

    2019

  48. [56]

    Leveraging evidential deep learning uncertainties with graph-based clustering to detect anomalies

    Sandeep Kumar Singh, Jaya Shradha Fowdur, Jakob Gawlikowski, and Daniel Medina. Leveraging evidential deep learning uncertainties with graph-based clustering to detect anomalies. IEEE Transactions on Intelligent Transportation Systems , 25, 2022

    2022

  49. [57]

    Uncertainty-aware deep classifiers using generative models

    Murat Sensoy, Lance Kaplan, Federico Cerutti, and Maryam Saleki. Uncertainty-aware deep classifiers using generative models. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 34, pages 5620--5627, 2020

    2020

  50. [58]

    Evidential deep learning to quantify classification uncertainty

    Murat Sensoy, Lance Kaplan, and Melih Kandemir. Evidential deep learning to quantify classification uncertainty. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors, Advances in Neural Information Processing Systems 31 , pages 3179--3189. 2018

    2018

  51. [59]

    Evidential Deep Learning to Quantify Classification Uncertainty

    Murat Sensoy, Lance Kaplan, and Melih Kandemir. Evidential Deep Learning to Quantify Classification Uncertainty . In NeurIPS , June 2018

    2018

  52. [60]

    Decision making in the TBM : The necessity of the pignistic transformation

    Philippe Smets. Decision making in the TBM : The necessity of the pignistic transformation. International Journal of Approximate Reasoning , 38(2):133--147, February 2005

    2005

  53. [61]

    Misclassification risk and uncertainty quantification in deep classifiers

    Murat Sensoy, Maryam Saleki, Simon Julier, Reyhan Aydogan, and John Reid. Misclassification risk and uncertainty quantification in deep classifiers. In IEEE Winter Conference on Applications of Computer Vision (WACV'21), Waikoloa, HI, USA, January 3-8, 2021 , pages 2483--2491....

    2021

  54. [62]

    Functional variational bayesian neural networks

    Shengyang Sun, Guodong Zhang, Jiaxin Shi, and Roger Grosse. Functional variational bayesian neural networks. In Proceedings of ICLR2019 , 2019

    2019

  55. [63]

    Dusenberry, Du Phan, Mark Patrick Collier, Jie Jessie Ren, Kehang Han, Zi Wang, Zelda Mariet, Clara Huiyi Hu, Neil Band, Tim G

    Dustin Tran, Jeremiah Liu, Michael W. Dusenberry, Du Phan, Mark Patrick Collier, Jie Jessie Ren, Kehang Han, Zi Wang, Zelda Mariet, Clara Huiyi Hu, Neil Band, Tim G. J. Rudner, Karan Singhal, Zachary Nado, Joost van Amersfoort, Andreas Christian Kirsch, Rodolphe Jenatton, Nith...

    2022

  56. [64]

    Prior and posterior networks: A survey on evidential deep learning methods for uncertainty estimation

    Dennis Thomas Ulmer, Christian Hardmeier, and Jes Frellsen. Prior and posterior networks: A survey on evidential deep learning methods for uncertainty estimation. Transactions on Machine Learning Research , April 2023

    2023

  57. [65]

    Uncertainty estimation using a single deep deterministic neural network

    Joost Van Amersfoort, Lewis Smith, Yee Whye Teh, and Yarin Gal. Uncertainty estimation using a single deep deterministic neural network. In International Conference on Machine Learning , pages 9690--9700. PMLR, 2020

    2020

  58. [66]

    Maximizing bci human feedback using active learning

    Zizhao Wang, Junyao Shi, Iretiayo Akinola, and Peter Allen. Maximizing bci human feedback using active learning. In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 10945--10951. IEEE, 2020

    2020

  59. [67]

    Uncertainty estimation for stereo matching based on evidential deep learning

    Chen Wang, Xiang Wang, Jiawei Zhang, Liang Zhang, Xiao Bai, Xin Ning, Jun Zhou, and Edwin Hancock. Uncertainty estimation for stereo matching based on evidential deep learning. Pattern Recognition , 124:108498, 2022

    2022

Pith tools

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