Pith. sign in

REVIEW 5 major objections 6 minor 79 references

Active and transfer learning with partially Bayesian neural networks for materials and chemicals

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Only the first hidden and output layers need to be Bayesian to match full Bayesian active-learning accuracy, and theory-pretrained priors make early exploration faster.

desk verdict A solid, reproducible empirical recipe for PBNN layer placement and theory-pretrained priors in active learning, with a real gap in validating the frozen-layer approximation before claiming full-BNN parity. read the letter →

arxiv 2501.00952 v2 pith:REZ5OYHX submitted 2025-01-01 cond-mat.dis-nn cond-mat.mtrl-sciphysics.data-an

Sarah I. Allec , Maxim Ziatdinov This is my paper · ORCID
classification cond-mat.dis-nncond-mat.mtrl-sciphysics.data-an
keywords activelearningBayesianneuralnetworksuncertaintyquantificationpartiallytransfermaterialsinformaticscheminformaticsHamiltonianMonteCarlo
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 asks whether a neural network must be fully Bayesian to guide active learning in materials and chemistry, and answers no. It shows that treating only the first hidden layer and the output layer as probabilistic, with all other weights frozen after stochastic-weight-averaged training, reproduces the accuracy, negative log predictive density, and 95% coverage of a fully Bayesian network across four molecular and materials benchmark datasets, at nearly a quarter of the computational cost. It further shows that initializing the Bayesian priors with weights pre-trained on theoretical calculations such as molecular dynamics or density functional theory makes active learning of experimental data more efficient, with the largest gains in the first tens of exploration steps. The practical upshot is that uncertainty-driven experimental exploration can be made affordable in the small-data, noisy-data regime typical of physical science.

What carries the argument

The load-bearing object is the partially Bayesian neural network (PBNN), defined by a two-stage training procedure in Algorithm 1: first a deterministic MLP is trained with stochastic weight averaging, then Hamiltonian Monte Carlo with the No-U-Turn Sampler is run only over a chosen subset of layers -- here the first hidden layer and the output layer, or one of the later hidden layers plus the output layer -- with priors centered on the pre-trained deterministic weights, while all other weights stay frozen. The quantity doing the work is the posterior predictive uncertainty $U^{post} = \frac{1}{N}\sum_{i=1}^N (y^*_i - \mu^{post})^2$, which combines weight-sampling variability with the observation-noise samples $\sigma_i$; the paper's claim is that this quantity computed over the small probabilistic subset faithfully approximates the full posterior predictive distribution of Eq. (4). The transfer-learning variant uses the same machinery, with the prior means set to the weights of a deterministic network pre-trained on simulated data and the prior width $\tau$ controlling how strongly the theory constrains the posterior.

What would settle it

A layer-ablation calculation on the four benchmark datasets would settle it: estimate the predictive variance contributed by each layer's weights by sampling only that layer at a time, and compare the sum against a full-network sampling. If the frozen deeper layers contribute a large share of the full Bayesian network's total predictive variance, the PBNN(0,4) uncertainty estimate is not faithful, and its comparable coverage would be a coincidence rather than a consequence of the method.

Watch

Extended reading notes

Core claim

The central discovery is architectural: where stochasticity is placed in a Bayesian neural network matters more than how much of the network is stochastic. With a five-layer MLP, the configuration PBNN(0,4) -- probabilistic first hidden layer plus probabilistic output layer -- tracks a fully Bayesian network's RMSE and uncertainty quality on ESOL, FreeSolv, steel fatigue, and conductivity datasets, while configurations with probabilistic later hidden layers (PBNN(1,4), PBNN(2,4), PBNN(3,4)) show unstable, oscillatory uncertainty calibration. The same favorable configuration also cuts overall computation by nearly a factor of four. The transfer-learning result is that priors centered on weights pre-trained on simulation data are a form of domain knowledge injection: they produce better-calibrated uncertainties and faster early progress than zero-centered priors, with the prior width $\tau$ acting as a tunable degree of trust in the theory.

Load-bearing premise

The comparison to a fully Bayesian network assumes that freezing all unsampled layers at their stochastic-weight-averaged values does not discard a substantial share of the total weight uncertainty, so that the posterior predictive uncertainty computed over the two probabilistic layers faithfully approximates the full posterior predictive distribution.

Editorial extensions

If this is right

  • Active learning with uncertainty-driven acquisition becomes practical for small, noisy materials and chemistry datasets: PBNN(0,4) delivers full-Bayesian-quality coverage at roughly a quarter of the compute.
  • Autonomous experimental workflows can run end-to-end without human oversight of the deterministic stage, since the MAP-prior penalty added in Algorithm 1 is designed to prevent overfitting before the sampling stage begins.
  • Theory-to-experiment transfer shortens the early exploration phase: pre-trained priors give lower RMSE and better calibrated coverage in the first tens of active learning steps than uninformative priors.
  • The prior width $\tau$ provides a single dial for trust in computational models: narrow priors hold the model close to theory, wider priors let experimental data take over, and the paper notes the possibility of relaxing $\tau$ dynamically as data accumulate.
  • The layer-position rule -- stochastic first hidden layer plus stochastic output layer -- gives a concrete default architecture for future PBNN applications to tabular scientific data.

Reading between the lines

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

  • The apparent layer-position effect suggests a testable hypothesis the authors only gesture at: the first hidden layer's stochasticity is what lets the model adapt its input representation under sparse data, while later-layer weights are tightly constrained by the data and add little to predictive uncertainty; if true, the same rule should transfer to convolutional or graph networks, with the analo
  • One way to make the prior-width trade-off fully adaptive would be a scheduled 'cooling' policy that starts with a tight theory-informed prior and widens $\tau$ as experimental data accumulate; the paper's two datasets hint at such a schedule but do not implement it.
  • Because the authors used one MLP architecture and one set of hyperparameters across all four datasets, PBNN(0,4) could plausibly serve as a default configuration for similar descriptor-based active-learning tasks, but this generalization is an extrapolation beyond the tested benchmark scope.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes partially Bayesian neural networks (PBNNs), in which a deterministic network is first trained with stochastic weight averaging and then HMC/NUTS sampling is applied only to selected layers, with the remaining weights frozen at their SWA values. It compares PBNN configurations against a fully Bayesian network on active learning for a nonstationary toy problem, two molecular datasets (ESOL, FreeSolv), and two materials datasets (Steel fatigue/NIMS, Conductivity/HTEM), using RMSE, NLPD, and 95% coverage. It further tests transfer learning by initializing the prior means with weights from a deterministic model pretrained on computational data, sweeping the prior width tau, on a noisy version of FreeSolv and a Materials Project/experimental bandgap dataset. The main claims are that PBNN(0,4) matches Full BNN accuracy and uncertainty at roughly one quarter of the computational cost, and that theory-pretrained priors improve early active-learning performance and calibration.

Significance. If the claims hold, the paper provides a practical and timely result: it offers a concrete rule of thumb for MLP architectures in active learning (make the first hidden layer and the output layer probabilistic), demonstrates that Bayesian inference over a subset of layers can provide useful uncertainty quantification, and shows a principled way to inject computational priors into active-learning workflows. The study is reproducible in principle: code and data are linked, five random seeds are used, R-hat diagnostics are reported, and the number of probabilistic-layer configurations is varied systematically. The consistency of the layer-placement trend across four datasets and a toy problem is a genuine strength. However, the key comparison to Full BNN and the practical cost claim require additional quantitative validation before the headline conclusions can be accepted.

major comments (5)
  1. [Section 2.1.1, Algorithm 1, Eq. (6)] The central parity claim is supported only by a conditional predictive distribution. After freezing the deterministic layers at theta_det, Eq. (6) computes p(y|x*, D, theta_det), not the full posterior predictive p(y|x*, D) = integral p(y|x*, theta, sigma) p(theta, sigma|D) dtheta dsigma. The paper does not quantify how much weight uncertainty is removed by this point-mass approximation. The R-hat histograms in Appendix 1 are computed on the sampled probabilistic weights only, so they cannot detect bias or variance loss in the frozen layers, and the NLPD and coverage metrics in Figures 4 and 5 are aggregate calibration scores that can coincide for distributions with different shapes in the high-uncertainty regions selected by the acquisition function. I request a quantitative check: for example, report the distribution of predictive variances (epistemic and aleatoric components) for PBNN versus Full BNN over the acquisition-selected inputs, coverage conditional on uncertainty bins, or the sensitivity of PBNN predictions to different deterministic SWA checkpoints. Without such a check, the headline claim that PBNN uncertainty estimates are comparable to those of fully Bayesian networks is not established.
  2. [Algorithm 1, lines 10 and 14] The prior for probabilistic layers is centered at theta_det obtained by training on the same dataset D used for the subsequent HMC/NUTS inference. This is a data-dependent (empirical-Bayes) prior, and the manuscript does not discuss it as such. If the Full BNN baseline uses standard zero-centered priors, the PBNN-versus-Full-BNN comparison conflates layer partialization with prior information. Please either use a prior that does not depend on the AL training targets, or compare PBNN against a Full BNN with an equally data-informed prior, and discuss the effect of double use of D.
  3. [Section 3.2, Figures 4(a) and 4(b)] The statement that active learning with PBNN achieves accuracy comparable to (ESOL) or better than (FreeSolv) that of standard deterministic ML models cites external MoleculeNet numbers [77]. Those numbers were obtained with different splits, features, preprocessing, and training protocols, and they are not directly comparable to the active-learning curves shown here. Please add a deterministic ML model trained and evaluated under the identical protocol, including the same random 5% initial subsets and the same RMSE computation, so that the accuracy claim is testable.
  4. [Section 3.2] The statement 'it decreased the overall computational time by nearly a factor of four' is a core practical claim but is not supported by any measurement in the paper. Please report wall-clock times (or relative training and sampling times) per active-learning iteration for PBNN(0,4), other PBNN configurations, and Full BNN, together with hardware and sampler settings, and specify whether the factor refers to total end-to-end time or MCMC time only.
  5. [Section 3.5, Figures 7 and 8] The transfer-learning experiments make the last two hidden layers and the output layer probabilistic, rather than the PBNN(0,4) configuration that Sections 3.1 to 3.3 identify as the best. The text gives no rationale for this change. Because the transfer-learning comparisons use a different probabilistic-layer configuration, it is unclear whether the observed benefits of theory-pretrained priors generalize to the recommended architecture or are specific to the last-layer setup. Please test transfer learning with PBNN(0,4) as well, or justify the alternative configuration.
minor comments (6)
  1. [Section 2.2.1] The word 'inverval' should be 'interval'.
  2. [Algorithm 1, line 15] As written, 'Sample weights theta_l ~ p(theta_l)' reads as drawing from the prior rather than from the posterior; please clarify that NUTS generates posterior samples from the specified prior and likelihood.
  3. [References] Reference [54] should cite Hoffman and Gelman, not Homan and Gelman; reference [61] has an author-name formatting error ('Alp Kucukelbir David M. Blei and Jon D. McAuliffe').
  4. [Appendix 1] There are typos in 'the the range' and 'traditonally'.
  5. [Section 3.5] The Noisy-FreeSolv dataset is described as experimental after synthetic noise is added; please clarify explicitly that the 'experimental' component is simulated so that readers do not mistake it for a real experimental dataset.
  6. [Figure 2] The notation PBNN(0,4) appears in the code snippet before it is defined in Section 3.1; please define the notation earlier or add a pointer.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central PBNN claims are benchmark results against external datasets and a Full BNN baseline, not derivations that reduce to their inputs.

full rationale

The paper's central claims are established by benchmarking, not by construction. Algorithm 1 defines PBNN training as deterministic SWA training followed by HMC/NUTS over a selected subset of layers, and Eqs. (5)-(7) compute the posterior predictive mean and uncertainty from the resulting MCMC samples; the claim that PBNN(0,4) approaches Full BNN accuracy and uncertainty is then tested on external datasets (FreeSolv, ESOL, NIMS, HTEM) and compared against a Full BNN baseline, so the conclusion is not encoded in the definitions. The prior width tau is swept over (0.1, 0.5, 1.0) and reported as a trade-off rather than fitted to the target curves. Self-citations ([7], [8], [76]) motivate the setting and provide baselines but are not load-bearing for the central comparison. The genuine weakness is empirical: freezing deterministic layers at theta_det makes Eq. (6) a conditional posterior, and the paper validates this approximation with aggregate R-hat statistics and visual or aggregate calibration metrics rather than a direct check of the frozen-weight posterior concentration; this is a validation gap or correctness risk, not a circular reduction, because the PBNN outputs are still compared with held-out data and an independent Full BNN baseline.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

This is an empirical benchmark, not a derivation, so the ledger records the hand-chosen hyperparameters and modeling assumptions that the conclusions rest on. The prior width tau is the main adjustable quantity in the transfer learning section and the conclusion about wider priors is read from a three-point sweep. The MLP hidden width and the 5 percent initial subset are fixed by hand without sensitivity tests. The axioms are: standard HMC/NUTS posterior sampling, the fidelity of freezing deterministic layers, predictive variance as acquisition signal, and the validity of weight-space transfer from simulation to experiment. No new entities are introduced.

free parameters (3)
  • Prior width tau = 0.1, 0.5, 1.0 (sweep)
    Chosen by hand to test the trade-off between trusting the theory prior and adapting to experimental data; the transfer learning conclusion that wider priors perform better is read off this sweep.
  • MLP hidden layer width = Not stated in text
    The paper fixes all hidden layers to equal width across datasets but never reports the value; the layer-placement conclusion may depend on it.
  • Initial training subset size = 5 percent of each dataset
    The active learning trajectories start from 5 percent random samples; the paper states trends are consistent but does not compare initial sizes.
assumptions (4)
  • standard math Standard Bayesian inference and HMC/NUTS machinery produce samples from the posterior p(theta_l, sigma | D) over the probabilistic subset.
    Invoked throughout Section 2.1 and Algorithm 1; the paper does not verify convergence of each chain beyond aggregated R-hat.
  • domain assumption Freezing deterministic layers at SWA point estimates yields a faithful approximation of the full posterior predictive uncertainty.
    Core to the PBNN claim in Section 2.1.1 and Algorithm 1; never proven, only tested on four datasets.
  • domain assumption Predictive variance U_post (Eq. 6) is a suitable acquisition signal for active learning exploration.
    Section 2.2 uses argmax of U_post as the acquisition function; the paper does not compare with other acquisition functions.
  • domain assumption Pretrained weights from computational models are meaningful prior means for experimental data in the same feature space.
    Section 3.5 transfer learning; assumes the simulation-to-experiment shift is mild enough that weight-space proximity transfers, without explicit covariate shift handling.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Active and transfer learning with partially Bayesian neural networks for materials and chemicals." pith.science (2026). https://pith.science/paper/REZ5OYHX

@misc{pith2026250100952,
  author       = {Pith},
  title        = {Pith review of: Active and transfer learning with partially Bayesian neural networks for materials and chemicals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/REZ5OYHX}},
  note         = {Machine review of arXiv:2501.00952}
}
read the original abstract

Active learning, an iterative process of selecting the most informative data points for exploration, is crucial for efficient characterization of materials and chemicals property space. Neural networks excel at predicting these properties but lack the uncertainty quantification needed for active learning-driven exploration. Fully Bayesian neural networks, in which weights are treated as probability distributions inferred via advanced Markov Chain Monte Carlo methods, offer robust uncertainty quantification but at high computational cost. Here, we show that partially Bayesian neural networks (PBNNs), where only selected layers have probabilistic weights while others remain deterministic, can achieve accuracy and uncertainty estimates on active learning tasks comparable to fully Bayesian networks at lower computational cost. Furthermore, by initializing prior distributions with weights pre-trained on theoretical calculations, we demonstrate that PBNNs can effectively leverage computational predictions to accelerate active learning of experimental data. We validate these approaches on both molecular property prediction and materials science tasks, establishing PBNNs as a practical tool for active learning with limited, complex datasets.

Figures

Figures reproduced from arXiv: 2501.00952 by the authors.

Figure 1
Figure 1. (a) Schematic illustration of Partially Bayesian Neural Network (PBNN) operation. First, we train a determin [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. (a) Schematic representation of the partially Bayesian MLP employed in this study. The model consists of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Active learning results for 1D toy dataset with non-stationary features. Evolution of predictions and uncertainty [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of Partially Bayesian Neural Networks (PBNNs) and fully Bayesian neural network (Full BNN) [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Comparison of Partially Bayesian Neural Networks (PBNNs) and fully Bayesian neural network (Full BNN) [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: (a) Schematic workflow of transfer learning in active learning: a deterministic model is first trained on [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Transfer learning with pre-trained PBNNs applied to noisy FreeSolv dataset. (a ) RMSE, NLPD, and coverage [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Transfer learning with pre-trained PBNNs applied to Bandgaps dataset. (a ) RMSE, NLPD, and coverage [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

79 extracted references · 50 canonical work pages

  1. [77]

    MoleculeNet: A Benchmark for Molecular Machine Learning

    Zhenqin Wu et al. MoleculeNet: A Benchmark for Molecular Machine Learning . 2018. arXiv: 1703.00564 [cs.LG]. URL: https://arxiv.org/abs/1703.00564

    arXiv 2018

  2. [1]

    Active learning with statistical models

    David A. Cohn, Zoubin Ghahramani, and Michael I. Jordan. “Active learning with statistical models”. In: J. Artif. Int. Res. 4.1 (1996), pp. 129–145. ISSN : 1076-9757

    work page 1996

  3. [2]

    Active Learning Literature Survey

    Burr Settles. Active Learning Literature Survey . Computer Sciences Technical Report 1648. University of Wisconsin–Madison, 2009. URL: http://axon.cs.byu.edu/~martinez/classes/778/Papers/settles. activelearning.pdf

    work page 2009

  4. [3]

    Active learning accelerates the discovery of high strength and high ductility lead-free solder alloys

    Bin Cao et al. “Active learning accelerates the discovery of high strength and high ductility lead-free solder alloys”. In: Materials & Design 241 (2024). ISSN : 02641275. DOI: 10.1016/j.matdes.2024.112921

    arXiv 2024

  5. [5]

    Benchmarking active learning strategies for materials optimization and discovery

    Alex Wang et al. “Benchmarking active learning strategies for materials optimization and discovery”. In:Oxford Open Materials Science 2.1 (2022). ISSN : 2633-6979. DOI: 10.1093/oxfmat/itac006

    doi:10.1093/oxfmat/itac006 2022

  6. [6]

    Small data machine learning in materials science

    Pengcheng Xu et al. “Small data machine learning in materials science”. In: npj Computational Materials 9.1 (2023). ISSN : 2057-3960. DOI: 10.1038/s41524-023-01000-z

    doi:10.1038/s41524-023-01000-z 2023

  7. [7]

    Bayesian Conavigation: Dynamic Designing of the Material Digital Twins via Active Learning

    B. N. Slautin et al. “Bayesian Conavigation: Dynamic Designing of the Material Digital Twins via Active Learning”. In: ACS Nano 18.36 (2024), pp. 24898–24908. ISSN : 1936-086X (Electronic) 1936-0851 (Linking). DOI: 10.1021/acsnano.4c05368. URL: https://www.ncbi.nlm.nih.gov/pubmed/39183496

    work page arXiv 2024

  8. [8]

    Bayesian Active Learning for Scanning Probe Microscopy: From Gaussian Processes to Hypothesis Learning

    M. Ziatdinov et al. “Bayesian Active Learning for Scanning Probe Microscopy: From Gaussian Processes to Hypothesis Learning”. In: ACS Nano 16.9 (2022), pp. 13492–13512. ISSN : 1936-086X (Electronic) 1936-0851 (Linking). DOI: 10.1021/acsnano.2c05303. URL: https://www.ncbi.nlm.nih.gov/pubmed/36066996

    work page arXiv 2022

Show all 79 references
  1. [9]

    Phase Stability Through Machine Learning

    Raymundo Arróyave. “Phase Stability Through Machine Learning”. In:Journal of Phase Equilibria and Diffusion 43.6 (2022), pp. 606–628. ISSN : 1547-7037 1863-7345. DOI: 10.1007/s11669-022-01009-9

    2022 doi

  2. [10]

    Data-driven analysis and prediction of stable phases for high-entropy alloy design

    I. Peivaste, E. Jossou, and A. A. Tiamiyu. “Data-driven analysis and prediction of stable phases for high-entropy alloy design”. In: Sci Rep 13.1 (2023), p. 22556. ISSN : 2045-2322 (Electronic) 2045-2322 (Linking). DOI: 10.1038/s41598-023-50044-0 . URL: https://www.ncbi.nlm.ni...

    2023

  3. [11]

    A comparative study of predicting high entropy alloy phase fractions with traditional machine learning and deep neural networks

    Shusen Liu et al. “A comparative study of predicting high entropy alloy phase fractions with traditional machine learning and deep neural networks”. In: npj Computational Materials 10.1 (2024). ISSN : 2057-3960. DOI: 10.1038/s41524-024-01335-1

    2024 doi

  4. [12]

    Exploring high thermal conductivity polymers via interpretable machine learning with physical descriptors

    Xiang Huang et al. “Exploring high thermal conductivity polymers via interpretable machine learning with physical descriptors”. In: npj Computational Materials 9.1 (2023). ISSN : 2057-3960. DOI: 10.1038/s41524- 023-01154-w

    2023 doi

  5. [13]

    Predicting lattice thermal conductivity via machine learning: a mini review

    Yufeng Luo et al. “Predicting lattice thermal conductivity via machine learning: a mini review”. In:npj Computa- tional Materials 9.1 (2023). ISSN : 2057-3960. DOI: 10.1038/s41524-023-00964-2

    2023 doi

  6. [14]

    Interpretable Machine Learning Model on Thermal Conductivity Using Publicly Available Datasets and Our Internal Lab Dataset

    Nikhil K. Barua et al. “Interpretable Machine Learning Model on Thermal Conductivity Using Publicly Available Datasets and Our Internal Lab Dataset”. In: Chemistry of Materials 36.14 (2024), pp. 7089–7100. ISSN : 0897- 4756 1520-5002. DOI: 10.1021/acs.chemmater.4c01696

    2024 doi

  7. [15]

    Finding Unprecedentedly Low-Thermal-Conductivity Half-Heusler Semiconductors via High-Throughput Materials Modeling

    Jesús Carrete et al. “Finding Unprecedentedly Low-Thermal-Conductivity Half-Heusler Semiconductors via High-Throughput Materials Modeling”. In: Physical Review X 4.1 (2014). ISSN : 2160-3308. DOI: 10.1103/ PhysRevX.4.011019

    2014

  8. [16]

    Prediction of glass transition temperature of oxide glasses based on interpretable machine learning and sparse data sets

    Chengcheng Liu and Hang Su. “Prediction of glass transition temperature of oxide glasses based on interpretable machine learning and sparse data sets”. In: Materials Today Communications 40 (2024), p. 109691. ISSN : 2352-

    2024

  9. [17]

    Data-driven machine learning prediction of glass transition temperature and the glass-forming ability of metallic glasses

    Jingzi Zhang et al. “Data-driven machine learning prediction of glass transition temperature and the glass-forming ability of metallic glasses”. In: Nanoscale 15 (45 2023), pp. 18511–18522. DOI: 10.1039/D3NR04380K. URL: http://dx.doi.org/10.1039/D3NR04380K

    2023 doi

  10. [18]

    Machine-Learning-Based Prediction of the Glass Transition Temperature of Organic Compounds Using Experimental Data

    G. Armeli, J. H. Peters, and T. Koop. “Machine-Learning-Based Prediction of the Glass Transition Temperature of Organic Compounds Using Experimental Data”. In:ACS Omega 8.13 (2023), pp. 12298–12309. ISSN : 2470-1343 (Electronic) 2470-1343 (Linking). DOI: 10.1021/acsomega.2c081...

    2023

  11. [19]

    Predicting glass transition temperature and melting point of organic compounds via machine learning and molecular embeddings

    Tommaso Galeazzo and Manabu Shiraiwa. “Predicting glass transition temperature and melting point of organic compounds via machine learning and molecular embeddings”. In: Environmental Science: Atmospheres 2.3 (2022), pp. 362–374. ISSN : 2634-3606. DOI: 10.1039/d1ea00090j

    2022 doi

  12. [20]

    Interpretable Machine Learning Framework to Predict the Glass Transition Temperature of Polymers

    M. J. Uddin and J. Fan. “Interpretable Machine Learning Framework to Predict the Glass Transition Temperature of Polymers”. In: Polymers (Basel) 16.8 (2024). ISSN : 2073-4360 (Electronic) 2073-4360 (Linking). DOI: 10.3390/polym16081049. URL: https://www.ncbi.nlm.nih.gov/pubmed...

    2024

  13. [21]

    Accurate prediction of dielectric properties and bandgaps in materials with a machine learning approach

    Yilin Hu et al. “Accurate prediction of dielectric properties and bandgaps in materials with a machine learning approach”. In: Applied Physics Letters 125.15 (Oct. 2024), p. 152905. ISSN : 0003-6951. DOI: 10 . 1063 / 5 . 0223890. eprint: https : / / pubs . aip . org / aip / ap...

    2024 doi

  14. [22]

    Machine learning dielectric screening for the simulation of excited state properties of molecules and materials

    S. S. Dong, M. Govoni, and G. Galli. “Machine learning dielectric screening for the simulation of excited state properties of molecules and materials”. In: Chem Sci 12.13 (2021), pp. 4970–4980. ISSN : 2041-6520 (Print) 2041-6539 (Electronic) 2041-6520 (Linking). DOI: 10.1039/d...

    2021

  15. [23]

    Delta Machine Learning for Predicting Dielectric Properties and Raman Spectra

    Manuel Grumet et al. “Delta Machine Learning for Predicting Dielectric Properties and Raman Spectra”. In: The Journal of Physical Chemistry C 128.15 (2024), pp. 6464–6470. ISSN : 1932-7455. DOI: 10.1021/acs.jpcc. 4c00886

    2024 doi

  16. [24]

    Machine learning and atomistic origin of high dielectric permittivity in oxides

    Y . Shimano, A. Kutana, and R. Asahi. “Machine learning and atomistic origin of high dielectric permittivity in oxides”. In: Sci Rep 13.1 (2023), p. 22236. ISSN : 2045-2322 (Electronic) 2045-2322 (Linking). DOI: 10.1038/ s41598-023-49603-2 . URL: https://www.ncbi.nlm.nih.gov/p...

    2023

  17. [25]

    Opportunities and Challenges for Machine Learning in Materials Science

    Dane Morgan and Ryan Jacobs. “Opportunities and Challenges for Machine Learning in Materials Science”. In: Annual Review of Materials Research 50.1 (2020), pp. 71–103. ISSN : 1531-7331 1545-4118. DOI: 10.1146/ annurev-matsci-070218-010015

    2020

  18. [26]

    Advances of machine learning in materials science: Ideas and techniques

    Sue Sin Chong et al. “Advances of machine learning in materials science: Ideas and techniques”. In: Frontiers of Physics 19.1 (2023). ISSN : 2095-0462 2095-0470. DOI: 10.1007/s11467-023-1325-z

    2023 doi

  19. [27]

    Explainable machine learning in materials science

    Xiaoting Zhong et al. “Explainable machine learning in materials science”. In: npj Computational Materials 8.1 (2022). ISSN : 2057-3960. DOI: 10.1038/s41524-022-00884-7

    2022 doi

  20. [28]

    Recent advances and applications of machine learning in solid-state materials science

    Jonathan Schmidt et al. “Recent advances and applications of machine learning in solid-state materials science”. In: npj Computational Materials 5.1 (2019). ISSN : 2057-3960. DOI: 10.1038/s41524-019-0221-0

    2019 doi

  21. [29]

    Confidence intervals for random forests: the jackknife and the infinitesimal jackknife

    Stefan Wager, Trevor Hastie, and Bradley Efron. “Confidence intervals for random forests: the jackknife and the infinitesimal jackknife”. In: Journal of Machine Learning Research 15.1 (Jan. 2014), pp. 1625–1651. ISSN : 1532-4435

    2014

  22. [30]

    On calibration of modern neural networks

    Chuan Guo et al. “On calibration of modern neural networks”. In: Proceedings of the 34th International Conference on Machine Learning - Volume 70. ICML’17. Sydney, NSW, Australia: JMLR.org, 2017, pp. 1321– 1330

    2017

  23. [31]

    FFMDFPA: A FAIRification Framework for Materials Data with No-Code Flexible Semi-Structured Parser and Application Programming Interfaces

    B. He et al. “FFMDFPA: A FAIRification Framework for Materials Data with No-Code Flexible Semi-Structured Parser and Application Programming Interfaces”. In: J Chem Inf Model 63.16 (2023). He, Bing Gong, Zhuming Avdeev, Maxim Shi, Siqi eng Research Support, Non-U.S. Gov’t 2023...

    2023

  24. [32]

    Data-Driven Materials Science: Status, Challenges, and Perspectives

    Lauri Himanen et al. “Data-Driven Materials Science: Status, Challenges, and Perspectives”. In: Advanced Science 6.21 (2019), p. 1900808. DOI: https : / / doi . org / 10 . 1002 / advs . 201900808. eprint: https : / / advanced . onlinelibrary . wiley . com / doi / pdf / 10 . 10...

    2019 doi

  25. [33]

    Enhancing materials property prediction by leveraging computational and experimental data using deep transfer learning

    D. Jha et al. “Enhancing materials property prediction by leveraging computational and experimental data using deep transfer learning”. In: Nat Commun 10.1 (2019). Jha, Dipendra Choudhary, Kamal Tavazza, Francesca Liao, Wei-Keng Choudhary, Alok Campbell, Carelyn Agrawal, Ankit...

    2019

  26. [34]

    Carl Edward Rasmussen and Christopher K. I. Williams. Gaussian Processes for Machine Learning . The MIT Press, Nov. 2005. ISBN : 9780262256834. DOI: 10.7551/mitpress/3206.001.0001 . URL: https: //doi.org/10.7551/mitpress/3206.001.0001

    2005 doi

  27. [35]

    Practical Bayesian Optimization of Machine Learning Algorithms

    Jasper Snoek, Hugo Larochelle, and Ryan P Adams. “Practical Bayesian Optimization of Machine Learning Algorithms”. In: Advances in Neural Information Processing Systems. Ed. by F. Pereira et al. V ol. 25. Curran Associates, Inc., 2012. URL: https://proceedings.neurips.cc/paper...

    2012

  28. [36]

    Surrogates: Gaussian Process Modeling, Design, and Optimization for the Applied Sciences

    Robert Gramacy. Surrogates: Gaussian Process Modeling, Design, and Optimization for the Applied Sciences. Mar. 2020. ISBN : 9780367815493. DOI: 10.1201/9780367815493

    2020 doi

  29. [37]

    Gaussian process regression for materials and molecules

    V olker L Deringer et al. “Gaussian process regression for materials and molecules”. In:Chemical Reviews 121.16 (2021), pp. 10073–10141

    2021

  30. [38]

    Manifold Gaussian Processes for regression

    Roberto Calandra et al. “Manifold Gaussian Processes for regression”. In: 2016 International Joint Conference on Neural Networks (IJCNN). 2016, pp. 3338–3345. DOI: 10.1109/IJCNN.2016.7727626

    2016

  31. [39]

    Deep Kernel Learning

    Andrew Gordon Wilson et al. “Deep Kernel Learning”. In:Proceedings of the 19th International Conference on Artificial Intelligence and Statistics. Ed. by Arthur Gretton and Christian C. Robert. V ol. 51. Proceedings of Machine Learning Research. Cadiz, Spain: PMLR, Sept. 2016,...

    2016

  32. [40]

    Stochastic variational deep kernel learning

    Andrew Gordon Wilson et al. “Stochastic variational deep kernel learning”. In: Proceedings of the 30th Interna- tional Conference on Neural Information Processing Systems. NIPS’16. Barcelona, Spain: Curran Associates Inc., 2016, pp. 2594–2602. ISBN : 9781510838819

    2016

  33. [41]

    Deep Kernel learning for reaction outcome prediction and optimization

    S. Singh and J. M. Hernandez-Lobato. “Deep Kernel learning for reaction outcome prediction and optimization”. In: Commun Chem 7.1 (2024), p. 136. ISSN : 2399-3669 (Electronic) 2399-3669 (Linking). DOI: 10.1038/ s42004-024-01219-x . URL: https://www.ncbi.nlm.nih.gov/pubmed/38877182

    2024

  34. [42]

    Deep kernel learning improves molecular fingerprint prediction from tandem mass spectra

    K. Duhrkop. “Deep kernel learning improves molecular fingerprint prediction from tandem mass spectra”. In: Bioinformatics 38.Suppl 1 (2022), pp. i342–i349. ISSN : 1367-4811 (Electronic) 1367-4803 (Print) 1367-4803 (Linking). DOI: 10.1093/bioinformatics/btac260 . URL: https://w...

    2022 doi

  35. [43]

    Deep kernel methods learn better: from cards to process optimization

    Mani Valleti et al. “Deep kernel methods learn better: from cards to process optimization”. In:Machine Learning: Science and Technology 5.1 (2024). ISSN : 2632-2153. DOI: 10.1088/2632-2153/ad1a4f

    2024 doi

  36. [44]

    The promises and pitfalls of deep kernel learning

    Sebastian W. Ober, Carl E. Rasmussen, and Mark van der Wilk. “The promises and pitfalls of deep kernel learning”. In: Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence. Ed. by Cassio de Campos and Marloes H. Maathuis. V ol. 161. Proceedings...

    2021

  37. [45]

    Bayesian Methods for Neural Networks and Related Models

    D. M. Titterington. “Bayesian Methods for Neural Networks and Related Models”. In: Statistical Science 19.1 (2004). ISSN : 0883-4237. DOI: 10.1214/088342304000000099

    2004 doi

  38. [46]

    Bayesian approach for neural networks—review and case studies

    J. Lampinen and A. Vehtari. “Bayesian approach for neural networks—review and case studies”. In: Neural Networks 14.3 (2001), pp. 257–274

    2001

  39. [47]

    Conference Paper

    Anh Nguyen, Jason Yosinski, and Jeff Clune.Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. Conference Paper. 2015. DOI: 0.1109/CVPR.2015.7298640

    2015

  40. [48]

    A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks

    Dan Hendrycks and Kevin Gimpel. A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks. Conference Paper. 2017

    2017

  41. [49]

    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. Conference Paper. 2017

    2017

  42. [50]

    Monte Carlo Sampling Methods Using Markov Chains and Their Applications

    W. K. Hastings. “Monte Carlo Sampling Methods Using Markov Chains and Their Applications”. In: Biometrika 57.1 (1970), pp. 97–109

    1970

  43. [52]

    A Conceptual Introduction to Hamiltonian Monte Carlo

    Michael Betancourt. A Conceptual Introduction to Hamiltonian Monte Carlo . 2018. arXiv: 1701 . 02434 [stat.ME]. URL: https://arxiv.org/abs/1701.02434. 15 Active and transfer learning with Bayesian neural networks for materials and chemicals

    2018 arXiv

  44. [53]

    Yes, but Did It Work?: Evaluating Variational Inference

    Yuling Yao et al. “Yes, but Did It Work?: Evaluating Variational Inference”. In: Proceedings of the 35th International Conference on Machine Learning. Ed. by Jennifer Dy and Andreas Krause. V ol. 80. Proceedings of Machine Learning Research. PMLR, Oct. 2018, pp. 5581–5590. URL...

    2018

  45. [54]

    The No-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo

    Matthew D. Homan and Andrew Gelman. “The No-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo”. In: J. Mach. Learn. Res. 15.1 (Jan. 2014), pp. 1593–1623. ISSN : 1532-4435

    2014

  46. [55]

    What uncertainties do we need in Bayesian deep learning for computer vision?

    Alex Kendall and Yarin Gal. “What uncertainties do we need in Bayesian deep learning for computer vision?” In: Proceedings of the 31st International Conference on Neural Information Processing Systems. NIPS’17. Long Beach, California, USA: Curran Associates Inc., 2017, pp. 558...

    2017

  47. [56]

    Weight Uncertainty in Neural Network

    Charles Blundell et al. “Weight Uncertainty in Neural Network”. In: Proceedings of the 32nd International Conference on Machine Learning. Ed. by Francis Bach and David Blei. V ol. 37. Proceedings of Machine Learning Research. Lille, France: PMLR, July 2015, pp. 1613–1622. URL:...

    2015

  48. [57]

    Laplace Redux - Effortless Bayesian Deep Learning

    Erik Daxberger et al. “Laplace Redux - Effortless Bayesian Deep Learning”. In: Advances in Neu- ral Information Processing Systems . Ed. by M. Ranzato et al. V ol. 34. Curran Associates, Inc., 2021, pp. 20089–20103. URL: https : / / proceedings . neurips . cc / paper _ files /...

    2021

  49. [58]

    R.M. Neal. Bayesian Learning for Neural Networks. Lecture Notes in Statistics. Springer New York, 2012. ISBN : 9781461207450. URL: https://books.google.com/books?id=LHHrBwAAQBAJ

    2012

  50. [59]

    Bayesian learning via stochastic gradient langevin dynamics

    Max Welling and Yee Whye Teh. “Bayesian learning via stochastic gradient langevin dynamics”. In:Proceedings of the 28th International Conference on International Conference on Machine Learning. ICML’11. Bellevue, Washington, USA: Omnipress, 2011, pp. 681–688. ISBN : 9781450306195

    2011

  51. [60]

    On the expressiveness of approximate inference in Bayesian neural networks

    Andrew Y . K. Foong et al. “On the expressiveness of approximate inference in Bayesian neural networks”. In: Proceedings of the 34th International Conference on Neural Information Processing Systems . NIPS ’20. Vancouver, BC, Canada: Curran Associates Inc., 2020. ISBN : 9781713829546

    2020

  52. [61]

    Variational Inference: A Review for Statisticians

    Alp Kucukelbir David M. Blei and Jon D. McAuliffe. “Variational Inference: A Review for Statisticians”. In: Journal of the American Statistical Association 112.518 (2017), pp. 859–877. DOI: 10.1080/01621459.2017. 1285773. eprint: https://doi.org/10.1080/01621459.2017.1285773 ....

    2017

  53. [62]

    Good Initializations of Variational Bayes for Deep Models

    Simone Rossi, Pietro Michiardi, and Maurizio Filippone. “Good Initializations of Variational Bayes for Deep Models”. In: Proceedings of the 36th International Conference on Machine Learning. Ed. by Kamalika Chaudhuri and Ruslan Salakhutdinov. V ol. 97. Proceedings of Machine L...

    2019

  54. [63]

    Do Bayesian Neural Networks Need To Be Fully Stochastic? 2023

    Mrinank Sharma et al. Do Bayesian Neural Networks Need To Be Fully Stochastic? 2023. arXiv: 2211.06291 [cs.LG]. URL: https://arxiv.org/abs/2211.06291

    2023 arXiv

  55. [64]

    Variational Bayesian Last Layers

    James Harrison, John Willes, and Jasper Snoek. Variational Bayesian Last Layers. 2024. arXiv: 2404.11599 [cs.LG]. URL: https://arxiv.org/abs/2404.11599

    2024 arXiv

  56. [65]

    Averaging Weights Leads to Wider Optima and Better Generalization

    Pavel Izmailov et al. Averaging Weights Leads to Wider Optima and Better Generalization . 2019. arXiv: 1803.05407 [cs.LG]. URL: https://arxiv.org/abs/1803.05407

    2019 arXiv

  57. [66]

    Empirical Frequentist Coverage of Deep Learning Uncertainty Quantification Procedures

    Benjamin Kompa, Jasper Snoek, and Andrew L. Beam. “Empirical Frequentist Coverage of Deep Learning Uncertainty Quantification Procedures”. In: Entropy 23.12 (Nov. 2021), p. 1608. ISSN : 1099-4300. DOI: 10. 3390/e23121608. URL: http://dx.doi.org/10.3390/e23121608

    2021 doi

  58. [67]

    How to evaluate uncertainty estimates in machine learning for regression?

    Laurens Sluijterman, Eric Cator, and Tom Heskes. “How to evaluate uncertainty estimates in machine learning for regression?” In: Neural Networks 173 (2024), p. 106203. ISSN : 0893-6080. DOI: https://doi.org/ 10.1016/j.neunet.2024.106203 . URL: https://www.sciencedirect.com/sci...

    2024

  59. [68]

    RDKit: Open-source cheminformatics: https://www.rdkit.org

    RDKit Contributors. RDKit: Open-source cheminformatics: https://www.rdkit.org

  60. [69]

    A general-purpose machine learning framework for predicting properties of inorganic materials

    Logan Ward et al. “A general-purpose machine learning framework for predicting properties of inorganic materials”. In: npj Computational Materials 2.1 (2016), pp. 1–7

    2016

  61. [70]

    FreeSolv: a database of experimental and calculated hydration free energies, with input files

    D. L. Mobley and J. P. Guthrie. “FreeSolv: a database of experimental and calculated hydration free energies, with input files”. In: J Comput Aided Mol Des 28.7 (2014), pp. 711–20. ISSN : 1573-4951 (Electronic) 0920-654X (Print) 0920-654X (Linking). DOI: 10.1007/s10822-014-974...

    2014

  62. [71]

    ESOL: Estimating Aqueous Solubility Directly from Molecular Structure

    John S. Delaney. “ESOL: Estimating Aqueous Solubility Directly from Molecular Structure”. In: Journal of Chemical Information and Computer Sciences 44.3 (2004), pp. 1000–1005. DOI: 10.1021/ci034243x. 16 Active and transfer learning with Bayesian neural networks for materials a...

    2004 doi

  63. [72]

    Exploration of data science techniques to predict fatigue strength of steel from composition and processing parameters

    A. Agrawal et al. “Exploration of data science techniques to predict fatigue strength of steel from composition and processing parameters”. In: Integr Mater Manuf Innov 3 (2014), pp. 90–108. DOI: 10.1186/2193-9772-3-8

    2014 doi

  64. [73]

    An open experimental database for exploring inorganic materials

    A. Zakutayev et al. “An open experimental database for exploring inorganic materials”. In:Sci Data 5 (2018), p. 180053. ISSN : 2052-4463 (Electronic) 2052-4463 (Linking). DOI: 10.1038/sdata.2018.53. URL: https: //www.ncbi.nlm.nih.gov/pubmed/29611842

    2018

  65. [74]

    Commentary: The Materials Project: A materials genome approach to accelerating materials innovation

    Anubhav Jain et al. “Commentary: The Materials Project: A materials genome approach to accelerating materials innovation”. In: APL Materials 1.1 (2013). ISSN : 2166-532X. DOI: 10.1063/1.4812323

    2013 doi

  66. [75]

    Predicting the Band Gaps of Inorganic Solids by Machine Learning

    Ya Zhuo, Aria Mansouri Tehrani, and Jakoah Brgoch. “Predicting the Band Gaps of Inorganic Solids by Machine Learning”. In: The Journal of Physical Chemistry Letters 9.7 (2018), pp. 1668–1673. DOI: 10.1021/acs. jpclett.8b00124

    2018 doi

  67. [76]

    Active Learning with Fully Bayesian Neural Networks for Discontinuous and Nonstationary Data

    Maxim Ziatdinov. Active Learning with Fully Bayesian Neural Networks for Discontinuous and Nonstationary Data. 2024. arXiv: 2405.09817 [cs.LG]. URL: https://arxiv.org/abs/2405.09817

    2024 arXiv

  68. [78]

    Inference from Iterative Simulation Using Multiple Sequences

    Andrew Gelman and Donald B. Rubin. “Inference from Iterative Simulation Using Multiple Sequences”. In: Statistical Science 7.4 (1992), pp. 457–472. DOI: 10.1214/ss/1177011136 . URL: https://doi.org/10. 1214/ss/1177011136

    1992

  69. [79]

    General Methods for Monitoring Convergence of Iterative Simulations

    Stephen P. Brooks and Andrew Gelman. “General Methods for Monitoring Convergence of Iterative Simulations”. In: Journal of Computational and Graphical Statistics 7.4 (1998), pp. 434–455. DOI: 10.1080/10618600. 1998.10474787. 17 Active and transfer learning with Bayesian neural...

    1998

  70. [4928]

    URL: https://www.sciencedirect

    DOI: https://doi.org/10.1016/j.mtcomm.2024.109691 . URL: https://www.sciencedirect. com/science/article/pii/S2352492824016726. 13 Active and transfer learning with Bayesian neural networks for materials and chemicals

    2024

  71. [5497]

    URL: https://proceedings.mlr.press/v97/rossi19a.html

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.