Pith. sign in

REVIEW 4 major objections 5 minor 52 references

JaxSGMC: Modular stochastic gradient MCMC in JAX

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read JaxSGMC packages stochastic-gradient MCMC samplers as modular JAX building blocks.

desk verdict A solid, genuinely useful JAX library for SG-MCMC with a weak empirical verification section; worth reviewing, but the authors should add quantitative sampler checks. read the letter →

arxiv 2505.11190 v1 pith:GUNIWVJ3 submitted 2025-05-16 stat.CO stat.APstat.ML

classification stat.COstat.APstat.ML
keywords JAXstochasticgradientMCMCBayesiandeeplearninguncertaintyquantificationSG-MCMCsamplersmodularsoftwareMarkovchainMonteCarloposteriorsampling
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 presents JaxSGMC, an application-agnostic JAX library that implements several state-of-the-art stochastic gradient MCMC (SG-MCMC) samplers and exposes standard SG-MCMC building blocks for composing custom samplers. Its aim is to lower the barrier to Bayesian deep learning by letting a user with an existing JAX model switch from stochastic optimization to posterior sampling without changing the model code. If the library works as described, uncertainty quantification becomes accessible for large-scale models, including newer samplers such as replica-exchange SGLD and AMAGOLD that existing JAX libraries did not offer. The two worked examples show a pSGLD sampler assembled from blocks matching a gold-standard HMC in linear regression, and giving uncertainty-aware predictions on CIFAR-10.

What carries the argument

The central object is the modular sampler architecture: each sampler is built from independent modules---potential.py (log-likelihood/prior potentials), data.py (jit-compatible DataLoaders), adaption.py (RMSProp and covariance preconditioners), integrator.py (Langevin diffusion, leapfrog with friction, OBABO), solver.py (accept/reject or unconditional sample processing), and scheduler.py (step size, burn-in, thinning). The mathematical object that carries the dynamics is the stochastic potential $U(\theta) \approx -\frac{N}{n}\sum_{i} \log p(y_i|x_i,\theta,\mathcal{M}) - \log p(\theta|\mathcal{M})$, whose mini-batch gradient drives each integrator; composing modules around this potential lets the same building blocks express different samplers.

What would settle it

Run one of the library's samplers on a tractable target with a closed-form posterior, such as a Gaussian linear model with known variance, and compare the MCMC sample variance and the posterior predictive coverage to the nominal level. A mismatch beyond Monte Carlo error, especially in the noise injected by the Langevin integrator or in replica-exchange swap acceptance, would show the implementation does not sample the stated distribution.

Watch

Extended reading notes

Core claim

On its own terms, JaxSGMC claims that the full variety of modern SG-MCMC algorithms can be organized into a small set of reusable modules---potential evaluation, data batching, preconditioner adaptation, integrators, solvers, and schedulers---and that composing these modules yields both standard samplers like (preconditioned) SGLD, SGHMC, and SGGMC and more recent schemes like replica-exchange SGLD and AMAGOLD under one API. The library is designed so that sampler components can be jit-compiled end-to-end, including the data loading, which keeps Bayesian sampling runtime-competitive with stochastic optimization. This is offered as a practical path for making SG-MCMC a drop-in alternative to optimization in JAX deep-learning workflows.

Load-bearing premise

The whole contribution rests on the implemented samplers being faithful to the algorithms they cite, so that the Markov chains they produce actually converge to the intended posterior; the paper's verification is qualitative or indirect, and a subtle implementation bug would silently bias the uncertainty estimates.

Editorial extensions

If this is right

  • A JAX user with a model already written as a function can call a high-level alias to sample the posterior instead of optimizing, with no model rewrite.
  • Samplers that require advanced building blocks, such as parallel tempering and amortized Metropolis-Hastings acceptance, become available through a common API rather than as stand-alone code.
  • Custom samplers can be assembled from the documented blocks, letting practitioners tailor proposals, preconditioners, and schedules to a problem.
  • Because data loading lives inside the jit-compiled loop, SG-MCMC inference can run at a cost comparable to stochastic optimization; the paper reports pSGLD training of 200 epochs on CIFAR-10 at the same order of magnitude as one full-batch HMC proposal.

Reading between the lines

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

  • If the modular decomposition is as clean as presented, SG-MCMC research could shift toward composing and benchmarking blocks rather than reimplementing full samplers, making new samplers easier to compare on identical data-loading and scheduling code.
  • A natural stress test the paper leaves implicit is quantitative convergence checking: running the library's samplers on targets with known posterior moments would let users verify the noise scaling and acceptance steps that the qualitative agreement plot does not settle.
  • The CIFAR-10 example suggests a direct extension: the same API could be used to ablate which building block (preconditioner, integrator, or swap schedule) most improves posterior coverage, a comparison the paper does not carry out.
  • The authors indicate that pSGLD may not fully explore posterior volume; a testable next step is whether AMAGOLD or replica-exchange samplers in this library reduce that gap on the same neural-network potential benchmark.
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

4 major / 5 minor

Summary. The paper presents JaxSGMC, a JAX library for stochastic gradient Markov chain Monte Carlo (SG-MCMC) that provides pre-built samplers (pSGLD, SGHMC, reSGLD, AMAGOLD, SGGMC) and a modular API for composing custom samplers from building blocks such as integrators, schedulers, and data loaders. The software is designed to be domain-independent, with end-to-end jit compilation and host-device data transfer. The paper demonstrates the library with two examples: a linear regression model using a custom-built pSGLD sampler, compared qualitatively to NumPyro HMC, and a CIFAR-10 image classification task using a pre-built pSGLD sampler, reporting classification accuracy and certainty-threshold behavior. The central claims are that JaxSGMC faithfully implements state-of-the-art SG-MCMC samplers and that its modular structure lowers the barrier to using and developing SG-MCMC methods.

Significance. If the implementation is faithful, JaxSGMC fills a practical gap by offering a modular, JAX-native SG-MCMC library that supports recently proposed samplers (reSGLD, AMAGOLD) not available in other JAX-based libraries, and it could accelerate adoption of Bayesian UQ in deep learning and physical modeling. The paper ships a public repository, documentation, and code listings, which are strengths for a software paper. However, the empirical verification of the central sampling-fidelity claim is currently weak, especially for the novel samplers, and one of the code listings appears to contain a substantive error. These issues are load-bearing for a library whose purpose is to draw samples from the correct posterior.

major comments (4)
  1. [Section 3.1] The validation of pSGLD against NumPyro HMC is qualitative: the text states that the distributions "agree reasonably well" without any numerical diagnostics. To substantiate the claim that the sampler targets the correct posterior, report quantitative measures such as Wasserstein or maximum mean discrepancy, effective sample size, or repeated-seed scatter plots.
  2. [Section 3.2] The CIFAR-10 example reports classification accuracy and certainty thresholds but provides no convergence diagnostics (e.g., trace plots, ESS, Gelman-Rubin) or posterior-fidelity checks. Ensemble classification accuracy alone does not demonstrate that the sampler draws from the intended posterior; add calibration or predictive-coverage checks, or explicitly label the example as an illustrative runtime/feature demonstration.
  3. [Section 2.3 / Section 3] The samplers highlighted as novel contributions relative to prior JAX libraries—replica exchange SG-MCMC (reSGLD) and AMAGOLD—are never empirically validated. A subtle implementation error, such as an incorrect swap-acceptance probability in reSGLD or an incorrect noise scaling in the AMAGOLD proposal, would bias the stationary distribution while remaining invisible in the current examples. Add synthetic-data experiments with known target posteriors for these samplers.
  4. [Listing 2, Section 3.1] The `log_prior` function returns `1 / jnp.exp(sample["log_sigma"])`, which is not the log-density of an exponential prior. For σ = exp(log_sigma) and an exponential(1) prior, the log-density is `-jnp.exp(sample["log_sigma"])` up to an additive constant. As written, the example defines a different potential than described in the text, compromising the illustrative linear regression example.
minor comments (5)
  1. [Section 4] The phrase "an domain-independent library" should be "a domain-independent library".
  2. [Figure 2 caption] The caption says "contour plots of Gaussians obtained from the Hamiltonian Monte Carlo (HMC) method"; clarify whether the Gaussians are fitted to HMC samples or derived analytically from the linear regression posterior.
  3. [Section 3.2] The statement that "the cost of the whole pSGLD training of 200 epochs is the same order of magnitude as the cost of generating a single sample with the full-batch HMC" is vague; specify hardware, the HMC hyperparameters (number of leapfrog steps, trajectory length), and the exact runtime comparison.
  4. [Section 4] The text refers to "Stochastic Variation Inference"; it should be "Stochastic Variational Inference".
  5. [Table 2] Table 2 lists the module as `solvers.py` while the text and listings refer to `solver.py`; make the module name consistent.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: the samplers are implementations of cited external algorithms checked against an independent NumPyro benchmark, and the few self-citations are not load-bearing.

full rationale

The paper's central claim is that the JaxSGMC library implements published SG-MCMC samplers and exposes modular building blocks. The derivation chain here is one of software engineering rather than equation-fitting: no parameter is fitted to data and then renamed as a prediction. The linear-regression example compares pSGLD samples against a gold-standard Hamiltonian Monte Carlo implementation in NumPyro, i.e., an independent external reference, and the CIFAR-10 example checks accuracy and certainty-based test-set performance rather than asserting a posterior identity. The cited algorithms (SGLD, pSGLD, SGHMC, reSGLD, AMAGOLD, SGGMC) are external publications; the paper does not derive those samplers from its own outputs. The only self-referential elements appear in the Impact section, where the authors cite their own prior applications of JaxSGMC as anecdotal evidence of utility. Those citations are not load-bearing for the central claim that the library implements the listed samplers; removing them would not weaken the software description or the examples. The skeptical concern that verification is qualitative (e.g., distributions 'agree reasonably well' without quantitative diagnostics, and no distributional check for reSGLD or AMAGOLD) is a correctness or validation-strength issue, not circularity: a weak or missing test does not make the implementation's claim equivalent to its inputs by construction. Accordingly, the circularity burden is low; the appropriate score is 1 for the minor, non-load-bearing self-citations, with no circular steps identified.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard SG-MCMC convergence theory and on the assumption that the library's implementations are correct. The examples introduce hand-picked hyperparameters, but these do not constitute fitted constants in a derivation; they are user choices that affect the demonstration. No new entities are postulated.

free parameters (6)
  • step_size_first = 0.05 (linear regression), 0.001 (CIFAR)
    Initial step size for the pSGLD sampler, chosen by hand; affects mixing and burn-in.
  • step_size_last = 0.001 (linear regression)
    Final step size after polynomial decay; hand-picked.
  • gamma = 0.33
    Exponent of the polynomial step size decay; hand-picked.
  • burn_in = 2000 (linear regression), 35100 (CIFAR)
    Number of discarded initial iterations; chosen by hand.
  • batch_size = 256 (CIFAR)
    Mini-batch size; chosen for GPU memory and gradient noise trade-off.
  • prior_std = 10 (CIFAR Gaussian prior)
    Standard deviation of the prior over weights; user-specified and influences the posterior.
assumptions (4)
  • domain assumption Discretized stochastic gradient Langevin dynamics converge to the posterior as the step size tends to zero with appropriately added noise.
    Justifies the samplers in Sections 2.1 and 2.3; cited to Welling and Teh (2011) and Ma et al. (2015).
  • domain assumption Observations in the dataset are conditionally independent given the model parameters, so the log-likelihood factorizes over data points.
    Assumed in Section 2.2 to define the mini-batch potential in eq. (4).
  • domain assumption JAX's host-callback API correctly and efficiently transfers data from DataLoaders into jit-compiled computations.
    Core design choice in Section 2.2; if this API is incompatible with all JAX transformations, end-to-end jit compilation and the architecture's claimed speed benefits fail.
  • domain assumption The implementations of reSGLD, AMAGOLD, and SGGMC follow the original papers' algorithms exactly, including swap acceptance and amortized Metropolis-Hastings steps.
    No external verification is provided for these samplers; the paper only shows pSGLD examples in detail.

how reviews work

0 comments
Cite this review

Pith. "Pith review of JaxSGMC: Modular stochastic gradient MCMC in JAX." pith.science (2026). https://pith.science/paper/GUNIWVJ3

@misc{pith2026250511190,
  author       = {Pith},
  title        = {Pith review of: JaxSGMC: Modular stochastic gradient MCMC in JAX},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GUNIWVJ3}},
  note         = {Machine review of arXiv:2505.11190}
}
read the original abstract

We present JaxSGMC, an application-agnostic library for stochastic gradient Markov chain Monte Carlo (SG-MCMC) in JAX. SG-MCMC schemes are uncertainty quantification (UQ) methods that scale to large datasets and high-dimensional models, enabling trustworthy neural network predictions via Bayesian deep learning. JaxSGMC implements several state-of-the-art SG-MCMC samplers to promote UQ in deep learning by reducing the barriers of entry for switching from stochastic optimization to SG-MCMC sampling. Additionally, JaxSGMC allows users to build custom samplers from standard SG-MCMC building blocks. Due to this modular structure, we anticipate that JaxSGMC will accelerate research into novel SG-MCMC schemes and facilitate their application across a broad range of domains.

Figures

Figures reproduced from arXiv: 2505.11190 by the authors.

Figure 1
Figure 1. Software architecture of JaxSGMC. The modules are ordered hierarchically from left to right, and the specificity of a module to the machine learning problem increases from top to bottom. by adequately annealing the simulation timestep ∆t → 0 and adding the right amount of noise to the stochastic gradient [13, 34]. More recent SG-MCMC schemes offer enhanced MH steps to sample from the unbiased distribution at finite … view at source ↗
Figure 2
Figure 2. Left: Sampled standard deviation σ parameters (blue) compared to the data￾generating value (orange). Middle and right: First and second (Middle) and third and fourth (right) components of weights wi sampled with the pSGLD scheme (blue scatter plot) compared to contour plots of Gaussians obtained from the Hamiltonian Monte Carlo (HMC) method (red) implemented in NumPyro [41, 12]. shows superior performance with large… view at source ↗
Figure 3
Figure 3. Obtained distributions of logits for five randomly selected images from the [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 46 canonical work pages

  1. [1]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, arXiv preprint arXiv:1810.04805 (2018)

  2. [2]

    Grigorescu, B

    S. Grigorescu, B. Trasnea, T. Cocias, G. Macesanu, A survey of deep learning techniques for autonomous driving, J. Field Robot. 37 (3) (2020) 362–386

  3. [3]

    Miotto, F

    R. Miotto, F. Wang, S. Wang, X. Jiang, J. T. Dudley, Deep learning for healthcare: review, opportunities and challenges, Brief. Bioinformatics 19 (6) (2018) 1236–1246

  4. [4]

    Raissi, P

    M. Raissi, P. Perdikaris, G. E. Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, J. Comput. Phys. 378 (2019) 686–707

  5. [5]

    No´ e, A

    F. No´ e, A. Tkatchenko, K. R. M¨ uller, C. Clementi, Machine Learning for Molecular Simulation, Annu. Rev. Phys. Chem. 71 (2020) 361–390. 16

  6. [6]

    Thaler, J

    S. Thaler, J. Zavadlav, Learning neural network potentials from exper- imental data via Differentiable Trajectory Reweighting, Nat. Commun. 12 (2021) 6884

  7. [7]

    Tossou, C

    P. Tossou, C. Wognum, M. Craig, H. Mary, E. Noutahi, Real-world molecular out-of-distribution: Specification and investigation, chemrxiv preprint (2023)

  8. [8]

    Arakelyan, R

    S. Arakelyan, R. J. Das, Y. Mao, X. Ren, Exploring distributional shifts in large language models for code analysis, arXiv preprint arXiv:2303.09128 (2023)

Show all 52 references
  1. [9]

    Efron, R

    B. Efron, R. J. Tibshirani, An introduction to the bootstrap, CRC press, 1994

  2. [10]

    J. Lei, M. G’Sell, A. Rinaldo, R. J. Tibshirani, L. Wasserman, Distribution-free predictive inference for regression, J. Am. Stat. Assoc. 113 (523) (2018) 1094–1111

  3. [11]

    Lakshminarayanan, A

    B. Lakshminarayanan, A. Pritzel, C. Blundell, Simple and scalable pre- dictive uncertainty estimation using deep ensembles, in: Advances in Neural Information Processing Systems, Vol. 30, Long Beach, CA, USA, Dec. 4–9, 2017, pp. 6405–6416

  4. [12]

    R. M. Neal, Handbook of Markov Chain Monte Carlo, 1st Edition, Chap- man and Hall/CRC, New York, USA, 2011, Ch. MCMC using Hamilto- nian Dynamics, pp. 139–188

  5. [13]

    Welling, Y

    M. Welling, Y. W. Teh, Bayesian learning via stochastic gradient Langevin dynamics, in: Proceedings of the 28th International Confer- ence on Machine Learning, Bellevue, WA, USA, Jun. 28 – Jul. 2, 2011, pp. 681–688

  6. [14]

    Graves, Practical variational inference for neural networks, in: Ad- vances in neural information processing systems, Vol

    A. Graves, Practical variational inference for neural networks, in: Ad- vances in neural information processing systems, Vol. 24, 2011

  7. [15]

    M. D. Hoffman, A. Gelman, The No-U-Turn Sampler: Adaptively Set- ting Path Lengths in Hamiltonian Monte Carlo, J. Mach. Learn. Res. 15 (2014) 1593–1623. 17

  8. [16]

    T. Chen, E. Fox, C. Guestrin, Stochastic gradient Hamiltonian Monte Carlo, in: Proceedings of the 31st International Conference on Machine Learning, Beijing, China, Jun. 21–26, 2014, pp. 1683–1691

  9. [17]

    C. Li, C. Chen, D. E. Carlson, L. Carin, Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks, in: Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence, Phoenix, AZ, USA, February 12–17, 2016, pp. 1788–1794

  10. [18]

    Nemeth, P

    C. Nemeth, P. Fearnhead, Stochastic gradient Markov chain Monte Carlo, J. Am. Stat. Assoc. 116 (533) (2021) 433–450

  11. [19]

    G. Lamb, B. Paige, Bayesian Graph Neural Networks for Molecular Property Prediction, in: Machine Learning for Molecules Workshop at NeurIPS, MIT Press, Online, Dec. 12, 2020

  12. [20]

    Z. Zou, X. Meng, A. F. Psaros, G. E. Karniadakis, NeuralUQ: A com- prehensive library for uncertainty quantification in neural differential equations and operators, arXiv preprint arXiv:2208.11866 (2022)

  13. [21]

    J. V. Dillon, I. Langmore, D. Tran, E. Brevdo, S. Vasudevan, D. Moore, B. Patton, A. Alemi, M. Hoffman, R. A. Saurous, Tensorflow distribu- tions, arXiv preprint arXiv:1711.10604 (2017)

  14. [22]

    Bingham, J

    E. Bingham, J. P. Chen, M. Jankowiak, F. Obermeyer, N. Pradhan, T. Karaletsos, R. Singh, P. A. Szerlip, P. Horsfall, N. D. Goodman, Pyro: Deep Universal Probabilistic Programming, J. Mach. Learn. Res. 20 (2019) 1–6

  15. [23]

    M. D. Hoffman, D. M. Blei, C. Wang, J. Paisley, Stochastic Variational Inference, J. Mach. Learn. Res. 14 (2013) 1303–1347

  16. [24]

    Baker, P

    J. Baker, P. Fearnhead, E. B. Fox, C. Nemeth, sgmcmc: An R Pack- age for Stochastic Gradient Markov Chain Monte Carlo, J. Stat. Softw. 91 (3) (2019) 1–27

  17. [25]

    A. K. Gupta, SG-MCMC (2016). URL https://github.com/akshaykgupta/SG_MCMC

  18. [26]

    Coullon, C

    J. Coullon, C. Nemeth, SGMCMCJax: a lightweight JAX library for stochastic gradient Markov chain Monte Carlo algorithms, J. Open Source Softw. 7 (72) (2022) 4113. 18

  19. [27]

    W. Deng, Q. Feng, L. Gao, F. Liang, G. Lin, Non-convex Learning via Replica Exchange Stochastic Gradient MCMC, in: Proceedings of the 37th International Conference on Machine Learning, PMLR, Online, Jul. 13–18, 2020, pp. 2474–2483

  20. [28]

    Zhang, A

    R. Zhang, A. F. Cooper, C. De Sa, AMAGOLD: Amortized Metropolis adjustment for efficient stochastic gradient MCMC, in: International Conference on Artificial Intelligence and Statistics, PMLR, Online, Aug. 26–28, 2020, pp. 2142–2152

  21. [29]

    Garriga-Alonso, V

    A. Garriga-Alonso, V. Fortuin, Exact Langevin Dynamics with Stochas- tic Gradients, in: 3rd Symposium on Advances in Approximate Bayesian Inference, Online, Jan. – Feb., 2021

  22. [30]

    Gallego, D

    V. Gallego, D. R. Insua, Stochastic Gradient MCMC with Repulsive Forces, in: Bayesian Deep Learning Workshop at NeurIPS, MIT Press, Montreal, Canada, Dec. 7, 2018

  23. [31]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas, NV, USA, Jun. 27–30, 2016, pp. 770–778

  24. [32]

    Krizhevsky, G

    A. Krizhevsky, G. Hinton, Learning multiple layers of features from tiny images, Tech. rep., University of Toronto (2009)

  25. [33]

    W. K. Hastings, Monte Carlo sampling methods using Markov chains and their applications, Biometrika 57 (1) (1970) 97–109

  26. [34]

    Y.-A. Ma, T. Chen, E. B. Fox, A Complete Recipe for Stochastic Gra- dient MCMC, in: Advances in Neural Information Processing Systems, Vol. 28, MIT Press, Montreal, Canada, 2015, p. 2917–2925

  27. [35]

    S. Kim, Q. Song, F. Liang, Stochastic gradient Langevin dynamics with adaptive drifts, J. Stat. Comput. Simul. 92 (2) (2022) 318–336

  28. [36]

    Zhang, C

    R. Zhang, C. Li, J. Zhang, C. Chen, A. G. Wilson, Cyclical Stochas- tic Gradient MCMC for Bayesian Deep Learning, in: 7th International Conference on Learning Representations, New Orleans, LA, USA, May 6–9, 2019. 19

  29. [37]

    Babuschkin, K

    I. Babuschkin, K. Baumli, A. Bell, S. Bhupatiraju, J. Bruce, P. Buchlovsky, D. Budden, T. Cai, A. Clark, I. Danihelka, C. Fan- tacci, J. Godwin, C. Jones, R. Hemsley, T. Hennigan, M. Hessel, S. Hou, S. Kapturowski, T. Keck, I. Kemaev, M. King, M. Kunesch, L. Martens, H. Merzic...

  30. [38]

    Tieleman, G

    T. Tieleman, G. Hinton, Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude, COURSERA: Neural networks for machine learning 4 (2) (2012) 26–31

  31. [39]

    S. Ahn, A. Korattikara, M. Welling, Bayesian Posterior Sampling via Stochastic Gradient Fisher Scoring, in: Proceedings of the 29th Inter- national Conference on Machine Learning, Omnipress, Madison, WI, USA, Jun. 26 – Jul. 1, 2012, pp. 1771–1778

  32. [40]

    Y. W. Teh, A. H. Thiery, S. J. Vollmer, Consistency and Fluctuations For Stochastic Gradient Langevin Dynamics, J. Mach. Learn. Res. 17 (2016) 1–33

  33. [41]

    D. Phan, N. Pradhan, M. Jankowiak, Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro, in: Program Transformations for ML at NeurIPS, MIT Press, Vancouver, Canada, Dec. 14, 2019

  34. [42]

    Hennigan, T

    T. Hennigan, T. Cai, T. Norman, I. Babuschkin, Haiku: Sonnet for JAX (2020). URL http://github.com/deepmind/dm-haiku

  35. [43]

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, H. Adam, MobileNets: Efficient Convolu- tional Neural Networks for Mobile Vision Applications, arXiv preprint arXiv:1704.04861 (2017)

  36. [44]

    J. Kim, S. Choi, Automated machine learning for soft voting in an en- semble of tree-based classifiers, in: International Workshop on Auto- matic Machine Learning at ICML, Stockholm, Sweden, Jul. 14, 2018. 20

  37. [45]

    Thaler, G

    S. Thaler, G. Doehner, J. Zavadlav, Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls, J. Chem. Theory Comput. (2023)

  38. [46]

    Wang, D.-Y

    H. Wang, D.-Y. Yeung, A survey on bayesian deep learning, ACM Com- put. Surv. 53 (5) (2020) 1–37

  39. [47]

    P. Ren, Y. Xiao, X. Chang, P.-Y. Huang, Z. Li, B. B. Gupta, X. Chen, X. Wang, A survey of deep active learning, ACM Comput. Surv. 54 (9) (2021) 1–40

  40. [48]

    A. G. Wilson, P. Izmailov, Bayesian Deep Learning and a Probabilis- tic Perspective of Generalization, in: Advances in Neural Information Processing Systems, Vol. 33, Online, Dec. 6–12, 2020, pp. 4697–4708

  41. [49]

    Y. Gal, Z. Ghahramani, Dropout as a bayesian approximation: Repre- senting model uncertainty in deep learning, in: International Conference on Machine Learning, PMLR, 2016, pp. 1050–1059

  42. [50]

    Hansen, P

    L. Hansen, P. Salamon, Neural Network Ensembles, IEEE Trans. Pat- tern Anal. Machine Intell. 12 (10) (1990) 993–1001

  43. [51]

    Thaler, M

    S. Thaler, M. Stupp, J. Zavadlav, Deep coarse-grained potentials via relative entropy minimization, J. Chem. Phys. 157 (2022) 244103

  44. [52]

    Thaler, J

    S. Thaler, J. Zavadlav, Uncertainty Quantification for Molecular Models via Stochastic Gradient MCMC, in: 10th Vienna Conference on Math- ematical Modelling, Vienna, Austria, Jul. 27–29, 2022, pp. 19–20. 21

Pith tools

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