Pith. sign in

REVIEW 2 major objections 3 minor 45 references

Fisher8: Stabilizing Neural Heteroscedastic Regression via Output-Layer Fisher Geometry

T0 review · 2 major / 3 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Fisher8, an output-layer natural-gradient rule, stabilizes heteroscedastic regression and improves calibration.

desk verdict A useful empirical recipe and a nice diagnostic, but the paper's KL trust-radius claim (Eq. 16) is unsupported: the per-point normalization in Eq. 15c does not control output displacements in a shared network. read the letter →

arxiv 2608.10374 v1 pith:DX5F27AA submitted 2026-08-11 cs.LG

Sumedh Vemuganti , Nickvash Kani This is my paper · ORCID
classification cs.LG MSC 62B1068T07
keywords heteroscedasticregressionnaturalgradientFisherinformationGaussiannegativelog-likelihooduncertaintycalibrationstabilizationuncertainty-awarerepresentationsKLtrustregion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that recurring instabilities in training neural networks to predict both mean and variance under Gaussian negative log-likelihood are not flaws of the loss itself but of stepping in Euclidean parameter space. It derives Fisher8, an output-layer update that replaces Euclidean gradient steps with natural-gradient steps measured by the Fisher information of each point's predicted Gaussian, then normalizes each point's corrected gradient before backpropagation. The claimed payoff is a stabilizer with no data-dependent hyperparameters beyond learning rate, an approximate KL trust radius for each update, and empirical gains in RMSE, NLL, and calibration across benchmarks. The paper further contends that several independently proposed stabilizers are partial versions of this same geometric correction.

What carries the argument

The central object is the $2\times 2$ Fisher information matrix of the network's per-point Gaussian predictive distribution, $F(\theta)=\operatorname{diag}(e^{-s},1/2)$, which turns Euclidean steepest descent into natural-gradient descent constrained by KL divergence. Fisher8 applies $F(\theta)^{-1}$ only to the output-layer gradients, yielding the reoriented updates $e^s\nabla_\mu \ell$ and $2\nabla_s \ell$, and then normalizes each point's update vector to unit norm before backpropagation; that normalization is what gives every point equal distributional mobility per step and produces the batch trust radius. The mechanism's work is to make the shared learning rate control approximate KL movement rather than raw parameter distance.

What would settle it

Measure the true KL divergence between the batch predictive distributions before and after a Fisher8 update on a network with overconfident predictions; if the actual KL regularly exceeds the claimed approximate upper bound by a large factor, the trust-radius argument does not control training.

Watch

Extended reading notes

Core claim

Fisher8 is the claim that jointly predicting mean and log-variance with the Gaussian NLL becomes stable if output-layer gradients are preconditioned by the per-point Fisher information $F(\theta)=\operatorname{diag}(e^{-s},\tfrac12)$ for $\theta=(\mu,s)$, $s=\ln\sigma^2$, reorienting the mean update to $e^s\nabla_\mu \ell$ and the log-variance update to $2\nabla_s \ell$, then L2-normalizing each point's preconditioned gradient before summing over the batch. The paper derives an approximate batch KL trust radius $\tfrac12 e^{-\min(s)}\eta^2 + \tfrac14\eta^2$ that bounds how far the predictive product distribution can move, and reads this bound as limiting the incentive to inflate variance to mask poor mean fits. Empirically, the update reaches better RMSE, NLL, and ECE than the $\beta$-NLL and Faithful baselines on eight UCI regression sets, a weak-lensing cosmology benchmark, and a rotated-MNIST representation task, while retaining downstream feature quality without large weight-space regularization.

Load-bearing premise

The stability claim rests on treating each point's normalized gradient as an independent move in its own local geometry and summing those moves; this is a heuristic rather than a consequence of the batch's Fisher metric.

Editorial extensions

If this is right

  • Training heteroscedastic regressors with plain SGD and Gaussian NLL becomes viable without beta reweighting, gradient severing, or tuned regularizer corridors.
  • The beta-NLL exponent, Faithful's Newton steps and trunk severing, and the regularizer corridor of prior work are reinterpreted as overlapping pieces of a single Fisher correction, so future stabilizers can be designed geometrically.
  • Fisher8-trained networks inherit second-order behavior: faster progress at conservative learning rates and sharper degradation at high learning rates, so learning rate is the single sensitive dial.
  • Uncertainty-aware feature spaces survive training: on rotated MNIST, Fisher8's penultimate-layer features support digit classification at roughly 80% accuracy even when the regression noise is class-independent, where baseline features collapse to 35-52%.
  • The approximate KL trust radius gives a concrete post-hoc readout of how far each batch update moves the predictive distribution, which can be monitored during training.

Reading between the lines

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

  • The local KL-variance diagnostic introduced in the paper could be used as a general early-warning metric: when feature activity persists but output-distribution KL variance stalls, further training on the same objective is unlikely to improve predictive uncertainty.
  • Because Fisher8 only touches the output layer, the same correction should transfer to other likelihood families whose per-point Fisher is small and diagonal, such as classification or count regression, though the paper does not test them.
  • The claimed reconciliation of beta-NLL and Faithful suggests a testable prediction: tuning beta-NLL's exponent or the Faithful severing point should be replaceable by Fisher8's normalization at matched learning-rate schedules, with no loss in calibration.
  • If the approximate KL bound is loose for overconfident networks, an adaptive learning-rate schedule that measures the actual batch KL after each update may extend Fisher8's stable range.
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

2 major / 3 minor

Summary. The paper proposes Fisher8, an output-layer training correction for heteroscedastic regression. For each point, the Gaussian NLL gradient is preconditioned by the exact 2x2 Fisher matrix of the Gaussian predictive distribution (Eq. 13). The batch rule normalizes each point's preconditioned gradient signal to unit L2 norm before backpropagation (Eq. 15c). The authors claim that this normalization provides an approximate KL trust radius for the batch update (Eq. 16), that several previously proposed stabilizers (beta-NLL, Faithful, and regularization-based approaches) converge on overlapping components of this geometric correction, and that, empirically, Fisher8 improves RMSE, NLL, and ECE on UCI regression, weak-lensing cosmology, and rotated MNIST representation learning.

Significance. If the claims held, Fisher8 would be a valuable, hyperparameter-light stabilizer for heteroscedastic regression, and the unification of prior stabilizers would be a useful conceptual contribution. The per-point Fisher derivation is correct, the method is simple and easily implementable, and the empirical evaluation is broad, spanning synthetic, UCI, cosmological, and representation-learning benchmarks. The paper's central theoretical contribution, however, is the approximate KL trust radius, and that claim is not supported by the actual shared-weight update. The empirical results may stand independently, but the advertised geometric interpretation needs substantial revision before the paper can be accepted.

major comments (2)
  1. [Section 3.3, Eq. (16)] The claim that the unit normalization in Eq. (15c) guarantees ||δμ||₂ = ||δs||₂ = η is not correct for the actual network update. For a linear output head μ_i = w_μᵀh_i + b_μ with shared weights, the change in the j-th predicted mean is δμ_j = -η/||g_μ||₂ Σ_i g_{μ,i}(h_jᵀh_i + 1), where g_μ is the vector of per-point natural gradients; unless the feature Gram matrix is the identity and the features sum to zero, the per-point displacements are not the normalized gradient signals. Consequently, the second-order KL estimate in Eq. (16) is not a bound on the actual movement of the predictive distribution, and the advertised 'approximate KL trust radius' is not established. The authors should either derive a correct bound for the shared-weight update or revise the claim to refer to the backpropagated signal rather than the predictive distributions.
  2. [Section 3.2, Eq. (15c)] The batch update rule is a heuristic aggregation of per-point natural gradients: it normalizes each point's gradient to unit norm and sums the signals, but this operation is not derived from the Fisher metric of the batch loss L_θ = (1/B)Σ_i ℓ_{θ_i}. The per-point Fisher geometry is exact only for the single-point objective; for a batch with shared weights, the output-layer Jacobians couple the points, and no per-point normalization implies a KL constraint on the joint update. This should be stated explicitly as a design choice, and the 'batch-level trust radius' language in Section 3.3 should be qualified accordingly.
minor comments (3)
  1. [Appendix D and Section 3.1 footnote] The tone of Appendix D ('we believe we in fact ate', 'peer reviewers have thoroughly enlightened us') and the McLaren footnote is not appropriate for a journal article; please rewrite these as neutral remarks or remove them.
  2. [Section 4.2.2, Table 2] The statement that 'Fisher8 achieves the best RMSE, NLL, and ECE on nearly every benchmark' is slightly overstated: in Table 2, Baseline-NLL has lower ECE on Concrete (0.024 vs 0.026) and lower ECE on Kin8nm (0.013 vs 0.017). Please qualify the statement.
  3. [Section 4.4, Table 4 and Appendix C.2] The main-text claim that 'other methods collapse to 35–52%' feature accuracy is incomplete when read against Appendix C.2: in the input-independent noise regime, the Collier et al. [2021] and Megerle et al. [2023] baselines achieve 79.46% and 80.96% feature accuracy, respectively. The comparison in the main text should either include these results or state clearly that the 35–52% range refers only to the methods listed in Table 4.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the update rule is derived from standard Fisher geometry, and all headline results are checked against external benchmarks.

full rationale

The paper's derivation chain is self-contained and non-circular. The per-point natural-gradient rule (Eqs. 4-13) follows from the Gaussian NLL and the exact 2x2 Fisher information matrix of the predictive Gaussian (Eq. 10), using the standard Amari (1998) expansion, which is external prior work rather than the authors' own imported result. The batch rule (Eq. 15c) applies this per-point preconditioning with L2 normalization as a stated heuristic; no fitted parameter is later renamed as a prediction, and the method introduces no data-dependent hyperparameters beyond learning rate. The claimed approximate KL trust radius (Eq. 16) is a post hoc bound, not a quantity fitted to data, so any question about whether it correctly describes the actual shared-weight update is a mathematical-correctness concern, not circularity. The unifying-lens discussion of beta-NLL, Faithful, and Fisher8 is a post hoc identification of overlapping update components and does not feed back into the construction of Fisher8. All headline empirical claims are benchmarked against external datasets (UCI, the FAIR Universe weak-lensing challenge, and rotated MNIST following Immer et al. 2023), so the central results are falsifiable outside the paper's own fitted values. Self-citation is not load-bearing: the cited priors (Seitzer et al., Stirn et al., Wong-Toi et al., Amari, Martens, Immer et al.) are independent external work, and no uniqueness theorem or ansatz is imported from the authors' own prior publications. The paper even discloses its own limitations (learning-rate sensitivity and below-nominal coverage), which further indicates that results are not protected by circular reasoning. No step in the derivation reduces by construction to its own inputs, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central derivation rests on the Gaussian likelihood assumption and the standard second-order KL approximation. No free parameters or new entities are introduced.

assumptions (3)
  • domain assumption The observed data are generated as y = mu(x) + epsilon with epsilon ~ N(0, sigma^2(x)) (Eq. 1), so the likelihood is Gaussian and the Fisher information is analytic.
    The entire derivation of the Fisher matrix in Eq. 10a depends on the Gaussian noise model; if the noise is non-Gaussian, the FIM and natural gradients differ.
  • standard math The KL divergence between successive predictive distributions can be approximated by its second-order Maclaurin expansion (Eq. 9), and the step sizes are small enough for this to hold.
    The natural gradient derivation relies on the standard second-order approximation of KL; this is a common assumption in natural gradient methods.
  • domain assumption The network architecture is expressive enough to represent the true mean and variance functions; failures are attributed to training dynamics, not capacity.
    The paper argues that since the same network can reach a good solution with modified training, the model class is expressive enough (Sec. 2.2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fisher8: Stabilizing Neural Heteroscedastic Regression via Output-Layer Fisher Geometry." pith.science (2026). https://pith.science/paper/DX5F27AA

@misc{pith2026260810374,
  author       = {Pith},
  title        = {Pith review of: Fisher8: Stabilizing Neural Heteroscedastic Regression via Output-Layer Fisher Geometry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DX5F27AA}},
  note         = {Machine review of arXiv:2608.10374}
}
read the original abstract

Training neural networks to jointly predict mean and uncertainty estimates from noisy observations can be unstable, prompting a series of independent stabilization efforts. We argue that these interventions highlight a common underlying issue where gradient steps are poorly aligned with the geometry of the loss landscape. To better align updates with local curvature, we derive Fisher8, an output-layer gradient correction that reorients and rescales updates using Fisher geometry rather than Euclidean geometry. Unlike past stabilizers, Fisher8 introduces no data-dependent hyperparameters beyond learning rate and admits an approximate KL trust radius between successive predictive distributions. We show that prior stabilizers converge on overlapping components of this geometric correction. Across multidimensional regression and representation-learning tasks, Fisher8 obtains superior likelihood--error tradeoffs, predicts calibrated uncertainty estimates, and learns rich uncertainty-aware feature spaces.

Figures

Figures reproduced from arXiv: 2608.10374 by the authors.

Figure 1
Figure 1. Conceptual summary of four approaches to uncertainty estimation in neural networks. Col 1: Joint mean–variance prediction with Gaussian NLL [Nix and Weigend, 1994]. Col 2: Variance-reweighted gradients via Beta-NLL [Seitzer et al., 2022]. Col 3: Faithful mean estimation using a separate MSE head without backpropagation through the shared trunk [Stirn et al., 2023]. Col 4: Over-parameterization with distinct regulari… view at source ↗
Figure 2
Figure 2. Col 1: Parameter space showing Euclidean circle (solid) and Fisher ellipse (dashed) with 8 sampled directions. Col 2: Gaussian PDFs resulting from Euclidean perturbations from the black reference N (0, 0.3 2 ). Col 3: Gaussian PDFs resulting from Fisher perturbations. Col 4: KL divergence from base distribution for each direction under both metrics. with partial derivatives: ∇µℓ = −e −s (y − µ), (4a) ∇sℓ = 1 2 − 1 2… view at source ↗
Figure 3
Figure 3. Row 1: predicted fits under input-independent noise (left) and input-dependent noise (right), with the Oracle shown in the rightmost column. Solid red/green lines are the mean predictions of Baseline-NLL/Fisher8; black is the ground-truth mean. Shaded regions show predicted ±2σ; dotted lines show ground-truth ±2σ. Row 2: log10 Jacobian variance of the learned feature space within a 0.5-radius neighborhood over train… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Pareto diagrams comparing four heteroscedastic regression methods across eight UCI benchmark datasets at six learning rates (0.001–0.1). Each panel shows normalized RMSE vs. normalized NLL, with normalization performed per-dataset using z-scores. Individual dataset res…
Figure 5
Figure 5. Figure 5: Pareto diagrams comparing four heteroscedastic regression methods across eight UCI benchmark datasets at six learning rates (0.001–0.1). Each panel shows normalized ECE vs. normalized RMSE, per-dataset. Individual dataset results shown as small markers; method means sh…
Figure 6
Figure 6. Figure 6: Zoomed convergence maps governed by the same cosmological parameters with different noise/realization [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: AI-generated image of a fictitious scenario in which a forager consults an AI chatbot about berry edibility. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 35 canonical work pages

  1. [1]

    Natural gradient works efficiently in learning

    Shun-ichi Amari. Natural gradient works efficiently in learning. Neural Computation, 10 0 (2): 0 251--276, 1998. doi:10.1162/089976698300017746

    doi:10.1162/089976698300017746 1998

  2. [2]

    Weight uncertainty in neural networks

    Charles Blundell, Julien Cornebise, Koray Kavukcuoglu, and Daan Wierstra. Weight uncertainty in neural networks. In International Conference on Machine Learning, pages 1613--1622. PMLR, 2015

    2015

  3. [3]

    Correlated input-dependent label noise in large-scale image classification

    Mark Collier, Basil Mustafa, Efi Kokiopoulou, Rodolphe Jenatton, and Jesse Berent. Correlated input-dependent label noise in large-scale image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1551--1560, 2021

    work page 2021

  4. [4]

    Laplace redux--effortless bayesian deep learning

    Erik Daxberger, Agustinus Kristiadi, Alexander Immer, Runa Eschenhagen, Matthias Bauer, and Philipp Hennig. Laplace redux--effortless bayesian deep learning. In Advances in Neural Information Processing Systems, volume 34, pages 20089--20103, 2021

    work page 2021

  5. [5]

    UCI machine learning repository, 2017

    Dheeru Dua and Casey Graff. UCI machine learning repository, 2017. URL http://archive.ics.uci.edu/ml

    2017

  6. [6]

    NeurIPS 2025 weak lensing uncertainty challenge

    FAIR Universe Collaboration . NeurIPS 2025 weak lensing uncertainty challenge. https://github.com/FAIR-Universe/Cosmology_Challenge, 2025. NeurIPS 2025 competition

    work page 2025

  7. [7]

    Dropout as a B ayesian approximation: Representing model uncertainty in deep learning

    Yarin Gal and Zoubin Ghahramani. Dropout as a B ayesian approximation: Representing model uncertainty in deep learning. In Maria Florina Balcan and Kilian Q. Weinberger, editors, Proceedings of the 33rd International Conference on Machine Learning, volume 48 of Proceedings of Machine Learning Research, pages 1050--1059. PMLR, 2016

    work page 2016

  8. [8]

    Practical variational inference for neural networks

    Alex Graves. Practical variational inference for neural networks. In Advances in Neural Information Processing Systems, volume 24, 2011

    work page 2011

Show all 45 references
  1. [9]

    Probabilistic backpropagation for scalable learning of bayesian neural networks

    Jos \'e Miguel Hern \'a ndez-Lobato and Ryan P Adams. Probabilistic backpropagation for scalable learning of bayesian neural networks. In International Conference on Machine Learning, pages 1861--1869, 2015

    2015

  2. [11]

    Alexander Immer, Emanuele Palumbo, Alexander Marx, and Julia E. Vogt. Effective bayesian heteroscedastic regression with deep neural networks. In Advances in Neural Information Processing Systems (NeurIPS), 2023

    2023

  3. [12]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In International Conference on Learning Representations (ICLR), 2015

    2015

  4. [13]

    Being bayesian, even just a bit, fixes overconfidence in relu networks

    Agustinus Kristiadi, Matthias Hein, and Philipp Hennig. Being bayesian, even just a bit, fixes overconfidence in relu networks. In International Conference on Machine Learning, pages 5436--5446. PMLR, 2020

    2020

  5. [14]

    Limitations of the empirical Fisher approximation for natural gradient descent

    Frederik Kunstner, Philipp Hennig, and Lukas Balles. Limitations of the empirical Fisher approximation for natural gradient descent. In Advances in Neural Information Processing Systems, volume 32, 2019

    2019

  6. [15]

    Simple and scalable predictive uncertainty estimation using deep ensembles

    Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell. Simple and scalable predictive uncertainty estimation using deep ensembles. In Advances in Neural Information Processing Systems, pages 6402--6413, 2017

    2017

  7. [16]

    New insights and perspectives on the natural gradient method

    James Martens. New insights and perspectives on the natural gradient method. Journal of Machine Learning Research, 21 0 (146): 0 1--76, 2020

    2020

  8. [17]

    Optimizing neural networks with kronecker-factored approximate curvature

    James Martens and Roger Grosse. Optimizing neural networks with kronecker-factored approximate curvature. In International Conference on Machine Learning, pages 2408--2417. PMLR, 2015

    2015

  9. [18]

    Stable optimization of G aussian likelihoods

    Denis Megerle, Fabian Otto, Michael Volpp, and Gerhard Neumann. Stable optimization of G aussian likelihoods. OpenReview preprint, 2023. URL https://openreview.net/forum?id=hmuLHC5MrG

    2023

  10. [19]

    Nix and Andreas S

    David A. Nix and Andreas S. Weigend. Estimating the mean and variance of the target probability distribution. In Proceedings of the 1994 IEEE International Conference on Neural Networks, volume 1, pages 55--60. IEEE, 1994

    1994

  11. [20]

    On the pitfalls of heteroscedastic uncertainty estimation with probabilistic neural networks

    Maximilian Seitzer, Arash Tavakoli, Dimitrije Antic, and Georg Martius. On the pitfalls of heteroscedastic uncertainty estimation with probabilistic neural networks. In International Conference on Learning Representations (ICLR), 2022

    2022

  12. [21]

    Sanjana, and David A

    Andrew Stirn, Hans-Hermann Wessels, Megan Schertzer, Laura Pereira, Neville E. Sanjana, and David A. Knowles. Faithful heteroscedastic regression with neural networks. In Proceedings of the 26th International Conference on Artificial Intelligence and Statistics. PMLR, 2023

    2023

  13. [22]

    Latent derivative bayesian last layer networks

    Joe Watson, Jihao Andreas Lin, Pascal Klink, Joni Pajarinen, and Jan Peters. Latent derivative bayesian last layer networks. In Proceedings of the 24th International Conference on Artificial Intelligence and Statistics, pages 1198--1206. PMLR, 2021

    2021

  14. [23]

    Understanding pathologies of deep heteroskedastic regression

    Eliot Wong-Toi, Alex Boyd, Vincent Fortuin, and Stephan Mandt. Understanding pathologies of deep heteroskedastic regression. In Proceedings of the 40th Conference on Uncertainty in Artificial Intelligence (UAI). PMLR, 2024

    2024

  15. [24]

    Nix and Andreas S

    David A. Nix and Andreas S. Weigend , title =. Proceedings of the 1994 IEEE International Conference on Neural Networks , year =

    1994

  16. [25]

    International Conference on Learning Representations (ICLR) , year =

    On the Pitfalls of Heteroscedastic Uncertainty Estimation with Probabilistic Neural Networks , author =. International Conference on Learning Representations (ICLR) , year =

  17. [26]

    Advances in Neural Information Processing Systems , pages =

    Simple and scalable predictive uncertainty estimation using deep ensembles , author =. Advances in Neural Information Processing Systems , pages =

  18. [27]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Effective Bayesian Heteroscedastic Regression with Deep Neural Networks , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  19. [28]

    Dropout as a

    Gal, Yarin and Ghahramani, Zoubin , booktitle =. Dropout as a. 2016 , editor =

    2016

  20. [29]

    Stable Optimization of

    Megerle, Denis and Otto, Fabian and Volpp, Michael and Neumann, Gerhard , year =. Stable Optimization of

  21. [30]

    Proceedings of the 26th International Conference on Artificial Intelligence and Statistics , year =

    Faithful Heteroscedastic Regression with Neural Networks , author =. Proceedings of the 26th International Conference on Artificial Intelligence and Statistics , year =

  22. [31]

    Proceedings of the 40th Conference on Uncertainty in Artificial Intelligence (UAI) , year =

    Understanding Pathologies of Deep Heteroskedastic Regression , author =. Proceedings of the 40th Conference on Uncertainty in Artificial Intelligence (UAI) , year =

  23. [32]

    International Conference on Machine Learning , pages =

    Optimizing Neural Networks with Kronecker-factored Approximate Curvature , author =. International Conference on Machine Learning , pages =. 2015 , organization =

    2015

  24. [33]

    Neural Computation , volume =

    Amari, Shun-ichi , title =. Neural Computation , volume =. 1998 , doi =

    1998

  25. [34]

    International Conference on Machine Learning , pages =

    Probabilistic backpropagation for scalable learning of Bayesian neural networks , author =. International Conference on Machine Learning , pages =

  26. [35]

    Advances in Neural Information Processing Systems , volume =

    Practical Variational Inference for Neural Networks , author =. Advances in Neural Information Processing Systems , volume =

  27. [36]

    International Conference on Machine Learning , pages =

    Weight Uncertainty in Neural Networks , author =. International Conference on Machine Learning , pages =. 2015 , organization =

    2015

  28. [37]

    International Conference on Machine Learning , pages =

    Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks , author =. International Conference on Machine Learning , pages =. 2020 , organization =

    2020

  29. [38]

    Advances in Neural Information Processing Systems , volume =

    Laplace Redux--Effortless Bayesian Deep Learning , author =. Advances in Neural Information Processing Systems , volume =

  30. [39]

    Journal of Machine Learning Research , volume =

    James Martens , title =. Journal of Machine Learning Research , volume =

  31. [40]

    Proceedings of the 24th International Conference on Artificial Intelligence and Statistics , pages =

    Latent Derivative Bayesian Last Layer Networks , author =. Proceedings of the 24th International Conference on Artificial Intelligence and Statistics , pages =. 2021 , publisher =

    2021

  32. [41]

    UCI Machine Learning Repository

    Dua, Dheeru and Graff, Casey. UCI Machine Learning Repository. 2017

    2017

  33. [42]

    2025 , howpublished =

    2025

  34. [43]

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

    Collier, Mark and Mustafa, Basil and Kokiopoulou, Efi and Jenatton, Rodolphe and Berent, Jesse , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages =

  35. [44]

    and Ba, Jimmy , title =

    Kingma, Diederik P. and Ba, Jimmy , title =. International Conference on Learning Representations (ICLR) , year =

  36. [45]

    arXiv preprint arXiv:2405.12807 , year =

    Hwang, Dongseong , title =. arXiv preprint arXiv:2405.12807 , year =

    arXiv

  37. [46]

    Advances in Neural Information Processing Systems , volume =

    Kunstner, Frederik and Hennig, Philipp and Balles, Lukas , title =. Advances in Neural Information Processing Systems , volume =

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.