Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural Networks

Cosmin Safta; Govinda Anantha Padmanabha; Nikolaos Bouklas; Reese E. Jones

REVIEW 5 major objections 5 minor 1 cited by

Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural Networks

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper introduces condensed Stein variational gradient descent, which prunes neural networks during training and gives posterior uncertainty on individual parameters, not just outputs.

desk verdict A promising but overclaimed method for concurrent sparsification and Stein UQ; the main fix is to either define the target the algorithm actually samples or stop calling it posterior UQ. read the letter →

arxiv 2412.16462 v1 pith:47DI2SJR submitted 2024-12-21 cs.LG physics.comp-phstat.ML

Govinda Anantha Padmanabha , Cosmin Safta , Nikolaos Bouklas , Reese E. Jones This is my paper · ORCID

classification cs.LGphysics.comp-phstat.ML

keywords BayesianneuralnetworksSteinvariationalinferencesparsificationgraphcondensationparameteruncertaintyquantificationinputconvexhyperelasticmaterialmodeling

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

cSVGD trains a neural network ensemble while pruning it and reporting uncertainties on individual weights, not just on predictions. The paper claims that treating each network in the ensemble as a graph, sorting nodes by an importance score, and forcing all graphs onto a shared, pruned template resolves the permutation ambiguity that normally makes parameter-level uncertainty meaningless. On a hyperelastic constitutive-modeling benchmark this reduces active parameters from 1020 to as low as 17 while keeping predictive accuracy, and makes each gradient-descent iteration cheaper as the shared graph shrinks. The authors present this as a concurrent alternative to their earlier sequential sparsify-then-infer approach, with the same accuracy at a fraction of the active parameter count.

What carries the argument

The central mechanism is the graph-condensation step (Algorithm 1) applied to a Stein ensemble: each network realization is treated as a layered directed graph, edges with weight below $\varepsilon$ are pruned, nodes are reordered by the importance score $s_{\ell,j}=\sum_i [W_\ell]_{ij}$, and all graphs are padded onto a common template whose layer widths are the per-layer maxima across the ensemble. The Stein flow itself combines three forces: a kernel-smoothed likelihood gradient that fits the data, a sparsifying prior term $\lambda|\theta|^\alpha$ that drives weights to zero, and a repulsive kernel term $\exp(-| \theta-\theta'|^\beta/\gamma^\beta)$ that keeps particles distinct. The graph condensation is what makes the sparsification stick: after each stage, near-zero weights cannot re-emerge because the particles evolve on the restricted template, and the common node ordering makes the parameter distributions across particles interpretable.

What would settle it

Run cSVGD on the same data and ensemble in two ways, once with the original parameterization and once with hidden nodes randomly permuted before each condensation step; if the final particle distributions or the held-out Wasserstein-1 distance change materially, then the importance-sorting reconciliation is not actually breaking the permutation symmetry.

Watch

Extended reading notes

Core claim

The paper develops a single variational inference procedure that sparsifies a feedforward neural network, fits it to data, and produces a particle ensemble whose per-parameter spread is a posterior uncertainty. It adds a sparsifying prior of the form $\pi(\theta)\propto\exp(-\lambda|\theta|^\alpha)$, $\alpha\in(0,1]$, to the Stein update, and interleaves the gradient flow with a graph-condensation step: prune edges below a threshold, rank nodes by their summed outgoing weight, and remap every ensemble member onto the largest common layer widths. The authors argue that this reconciliation removes spurious repulsion between equivalent permuted parameterizations, so the condensed ensemble approximates the original posterior while each particle uses far fewer active weights. In the mechanics test case, graph condensation lowers the final active parameter count from 1020 to between 17 and 127 depending on $\alpha$, with no loss in Wasserstein-1 predictive accuracy and an iteration cost that scales with the reduced parameter count.

Load-bearing premise

The method assumes that pruning small edge weights and sorting hidden nodes by their summed outgoing weight makes different networks in the ensemble genuinely comparable, so that forcing every particle onto one shared sparse graph does not distort the posterior the ensemble represents.

Editorial extensions

If this is right

For overparameterized networks with many negligible weights, cSVGD reduces active parameters from 1020 to as few as 17 in the demonstration while holding predictive accuracy fixed.
Because per-iteration cost scales with the number of active parameters, the condensed flow runs faster as training proceeds, more than offsetting the cost of sorting and remapping graphs.
The aligned ensemble yields posterior distributions on individual weights and nodes, so uncertainty can be attributed to specific connections rather than only to network outputs.
The adaptive penalty scheme reaches near-optimal accuracy-versus-sparsity trade-offs across noise levels, suggesting the method does not require delicate prior tuning.

Reading between the lines

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

The same importance-sorting reconciliation could be applied to other permutation-symmetric architectures, such as transformer attention heads, though the paper only demonstrates feedforward networks.
The condensation is a heuristic with no proven error bound; before using the resulting parameter variances for safety-critical decisions, one could validate them against full-ensemble Stein inference or Hamiltonian Monte Carlo on small problems.
The column-sum importance metric is only one possible ranking; ranking nodes by weight variance or by local loss sensitivity would likely produce different common graphs and different accuracy-sparsity frontiers.
If the method generalizes, a useful test would be calibration of the push-forward predictive intervals, since the paper reports Wasserstein distances but not coverage of held-out test points.

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.

Desk Editor's Note

read the letter

cSVGD is a genuine algorithmic contribution—running SVGD concurrently with sparsifying priors and a graph-condensation step to align the ensemble—but the paper overstates the parameter-UQ guarantee. As written, the algorithm changes its target during the run, so Eq. (25)'s claim that the particles approximate π(θ|D) does not follow.

What's new: combining sparsification and Stein flow at the same time (rather than sequentially as in the authors' Ref. [7]), plus the graph reconciliation that sorts nodes by importance and prunes to a common template. That permutation-alignment idea is simple and addresses a real problem in SVGD ensembles of neural networks. The experiments on the hyperelastic ICNN show the condensation does what it claims: parameter count drops from 1020 to tens of parameters with W1 accuracy comparable to their sequential baseline, and Fig. 8's with/without condensation comparison is a nice ablation. The writing is clear and the limitations are partly acknowledged.

The soft spots are real, and one is load-bearing. The central UQ claim requires a fixed target π(θ|D). But Alg. 2 adapts λ and then reverts to minimal λ; Alg. 1 prunes and pads the graph; and Sec. 4 says kernel effects are masked near parameter axes. Each of these changes the drift and repulsion in the Stein flow, so the final ensemble is not a sample from any well-defined posterior of the original network. The KDEs in Fig. 7 and the W1 pushforward intervals are presented as posterior uncertainty, but there's no argument connecting them to a fixed posterior. This is not a minor omission—it's the paper's main promise.

Other issues are smaller but worth naming: no code or data, single runs without seed variance, speed-up inferred from parameter counts rather than wall-clock time, and the only baseline is the authors' own sequential method. The condensation heuristic and the axis-masking rule are also underspecified. These are fixable, but they need to be addressed.

Who should read it: researchers working on Bayesian UQ for physics-based neural networks. It's a useful idea, honestly presented, and with revision could be a solid contribution. I'd send it to review, but with a clear message that the target-distribution problem and reproducibility are not optional.

Referee Report

5 major / 5 minor

Summary. The paper proposes condensed Stein variational gradient descent (cSVGD), an SVGD variant that concurrently sparsifies and trains a neural-network ensemble while targeting parameter-level uncertainty quantification. The method adds an alpha-exponential sparsifying prior, an adaptive penalty lambda, and a graph-condensation step (Alg. 1) that prunes, sorts, and reconciles the particles onto a common graph before each SVGD stage (Alg. 2). A modification that masks kernel effects near parameter axes is mentioned in Sec. 4. The authors demonstrate the method on a 3D Gaussian illustration and on an input-convex neural network for hyperelastic constitutive modeling, comparing sparsity and W1 pushforward accuracy with and without condensation and with a prior sequential L0/L2+Stein method.

Significance. If the claims hold, cSVGD would be a practically useful contribution: it addresses permutation symmetry in Bayesian neural-network ensembles, produces parameter-level UQ, and reduces model complexity while preserving predictive accuracy. The paper's strengths include a clear problem motivation, systematic sweeps over prior order, kernel bandwidth, ensemble size, and noise level (Figs. 3, 9-11), and an explicit comparison against the authors' previous sequential method (Fig. 12b). The main limitation is that the central UQ claim is not backed by a well-defined target posterior or by calibration experiments, and the empirical claims lack replication and wall-clock timing. No code or data is provided, which further limits verification.

major comments (5)

[Sec. 3.4, Eq. (25)] The central claim that the converged particles approximate pi(theta|D) is not supported because Alg. 2 changes the sampling target during the run. The SVGD derivation in Eqs. (11)-(15) assumes a fixed posterior pi and fixed kernel kappa, but the algorithm increases lambda and later reverts it, applies graph condensation via Alg. 1, and (per Sec. 4) masks kernel effects near parameter axes. No argument or theorem shows that the terminal ensemble is a sample from a well-defined posterior of the original network; please provide such an argument or, failing that, replace Eq. (25) with a more limited claim and validate the parameter-level UQ empirically (e.g., credible-interval coverage on synthetic data with known weights).
[Sec. 5.4, Fig. 8] The speed-up claim is not supported by wall-clock measurements. The text asserts that computational cost per iteration is roughly proportional to parameter count, but no runtimes are reported, and the overhead of graph sorting/reconfiguration is not quantified. Please report actual CPU/GPU wall-clock time for the experiments in Fig. 8 and for the comparison in Fig. 12b, including the condensation overhead. Additionally, Fig. 8 appears to show a single run per prior order; without multiple seeds or error bars, the claim that condensation gives 'no reduction in accuracy' across a wide range of priors is not statistically established.
[Sec. 3.3, Alg. 1] The graph-reconciliation heuristic is a load-bearing but unproven component. Sorting by the column-sum importance of Eq. (24) and pruning by |W|<epsilon is not shown to align permuted-but-equivalent parameterizations across the ensemble or to preserve the support of the posterior. Because the SVGD repulsion term depends on distances between particles, misalignment will distort the stationary ensemble. Please provide a formal statement of what Alg. 1 guarantees, or a synthetic experiment with known permutation symmetries that measures alignment error and shows that cSVGD's posterior (or predictive) distribution matches a reference SVGD/HMC solution.
[Sec. 4] The modification to 'mask kernel effects when the particles are close to parameter axes' is unspecified and is not part of the pseudocode in Alg. 2. This makes the experiments in Sec. 5 irreproducible and changes the kernel in Eq. (15) in a way not reflected in any target distribution. Please specify the exact masking rule (which axes, what threshold, how the mask is tapered) and, if it is essential, analyze or empirically test its effect on the stationary distribution.
[Sec. 3.2, Eqs. (16)-(18)] The kernel gradient equations are inconsistent as written: the derivative of kappa(theta',theta)=exp(-|theta'-theta|^beta/gamma^beta) is not 1/gamma |theta-theta'|^(beta-1) sign(theta-theta') kappa(theta',theta) as in Eq. (17), and Eq. (18) uses exp(-gamma|theta-theta'|^beta) rather than the kernel in Eq. (16). Since Eq. (15) relies on these gradients, the algorithm as specified is not implementable. Please correct the formulas and state the exact kernel used in the experiments.

minor comments (5)

[Sec. 3.1, Eqs. (6)-(7)] The density in Eq. (6) is not normalized for alpha != 1 with the constants in Eq. (7); for the intended exponential-power form it integrates to lambda^(1-1/alpha) alpha^(-1/alpha). This does not affect the score, but the claim that pi(theta) is a normalized prior should be corrected or the constants adjusted.
[Sec. 3.2, Eq. (12)] The index notation is inconsistent: the update writes theta_a^(k+1) but the gradient argument is g(theta_(i)^(k)); use the same particle index throughout.
[Sec. 5.3, Eq. (38)] The W1 distance definition omits the time/index dependence; since the text says the distances are summed across all output times, Eq. (38) should carry a time index or the summation should be made explicit.
[Alg. 1] The termination condition 'while ensemble is not changing topologically' is not formal; add a maximum number of condensation stages or a precise convergence criterion so that the algorithm is well-defined.
[General] No code or data availability statement is provided; given the number of heuristics and hyperparameters, releasing code and seed values would be important for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular dependency: the central derivation follows the standard SVGD update with an explicit prior, and the empirical claims are benchmarked against a synthetic truth model.

full rationale

The paper's derivation chain is self-contained rather than circular. The posterior target is defined by Bayes rule (Eq. 1) with an explicit likelihood (Eq. 2) and sparsifying prior (Eqs. 3 and 6), and the particle update (Eqs. 12-15) is the standard SVGD gradient derived from the Stein identity and the KL-divergence argument in Eq. 11; no quantity entering the update is fitted to the benchmark. The central empirical claims in Secs. 5.4-5.6 compare cSVGD with and without graph condensation, and against the authors' prior sequential method, using a synthetic truth model (Eq. 35) with known constants and controlled noise (Eq. 36). Accuracy is measured by Wasserstein distance to the pushforward of that known truth distribution, so the reported accuracy and sparsity results are externally anchored rather than forced by construction. The graph condensation procedure (Alg. 1) is explicitly heuristic, and the adaptive penalty and kernel masking in Alg. 2 mean the terminal particle ensemble may not correspond to a well-defined fixed posterior, so the Eq. (25) posterior-interpretation claim carries a correctness caveat; however, that is an unsupported claim about the method's target, not a circular reduction of a prediction to its inputs. The only self-references are to prior work [7] as a comparator and as the source of the synthetic data generator and ICNN architecture; neither is used to define the predicted output or the sparsity metric, and the comparison is against data generated independently of the fitted model. Therefore no specific circular step can be exhibited, and the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The method introduces no new physical entities. It relies on standard SVGD theory, the choice of exponential-family priors, and the ICNN architecture. The load-bearing heuristic is the graph condensation (Alg. 1): sorting by column sums (Eq. 24) and pruning with threshold epsilon = 1e-3 is assumed to align fungible parameters across the ensemble without biasing the posterior. The kernel-masking modification near parameter axes, mentioned in Sec. 4, is an additional unquantified assumption. Key hyperparameters (lambda, alpha, beta, gamma, epsilon, Nr, step size) are chosen by hand and some are not reported.

free parameters (6)

sparsifying prior multiplier lambda = 0.001, 0.01, 0.05, 0.1; fixed 0.05 in Sec. 5.5
Chosen by hand to balance sparsity and accuracy; adaptive scheme increases it until accuracy suffers.
prior order alpha = 0.25, 0.5, 0.75, 1, 2
Hyperparameter controlling sparsity shape; explored via sweeps and fixed per experiment.
kernel order beta = 2 for the application
Only beta = 2 kernel used after initial sweep; choice affects repulsion strength.
kernel bandwidth gamma = gamma = sqrt(d_median / (2 log(Nr + 1)))
Adapted using median interparticle distance (Silverman-like heuristic); not reported as a fixed value.
pruning threshold epsilon = 1e-3
Weights below this threshold are excluded in graph condensation; chosen by hand without sensitivity analysis.
kernel masking tolerance near axes = not specified
Sec. 4 modifies Alg. 2 to mask kernel effects when particles are close to parameter axes, but no numeric tolerance is given.

assumptions (7)

standard math Stein identity and the SVGD update rule hold for the target posterior (from Liu and Wang [4])
The algorithm relies on the standard SVGD derivation in Sec. 3.2.
domain assumption Gaussian likelihood correctly models the data noise
Eq. (2) assumes Gaussian noise with variance sigma^2; used for both toy and material examples.
domain assumption The exponential-family sparsifying prior in Eq. (6) is a suitable prior
This prior family induces sparsity but is chosen; the constants c1 and c2 are normalization factors, not derived from data.
domain assumption Input convex neural network with non-negative weights guarantees convexity of the output
Sec. 5.1 relies on the ICNN construction of Amos et al. [34] to enforce polyconvexity.
domain assumption The truth model in Eq. (35) and multiplicative noise in Eq. (36) define a valid test problem
Synthetic data generation follows previous work Ref. [7]; the ground-truth potential is assumed realistic for the demonstration.
ad hoc to paper Sorting nodes by column-sum importance (Eq. 24) achieves quasi-optimal edge similarity across the ensemble
Alg. 1 states this as a result but it is a heuristic with no proof; the alignment quality directly affects parameter UQ.
ad hoc to paper Pruning edges with |W| < epsilon and removing disconnected nodes preserves the posterior support
The condensation step discards parameters without showing that the target posterior is preserved; this is load-bearing for the sparsity claim.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural Networks." pith.science (2026). https://pith.science/paper/47DI2SJR

@misc{pith2026241216462,
  author       = {Pith},
  title        = {Pith review of: Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/47DI2SJR}},
  note         = {Machine review of arXiv:2412.16462}
}

read the original abstract

We propose a Stein variational gradient descent method to concurrently sparsify, train, and provide uncertainty quantification of a complexly parameterized model such as a neural network. It employs a graph reconciliation and condensation process to reduce complexity and increase similarity in the Stein ensemble of parameterizations. Therefore, the proposed condensed Stein variational gradient (cSVGD) method provides uncertainty quantification on parameters, not just outputs. Furthermore, the parameter reduction speeds up the convergence of the Stein gradient descent as it reduces the combinatorial complexity by aligning and differentiating the sensitivity to parameters. These properties are demonstrated with an illustrative example and an application to a representation problem in solid mechanics.

Figures

Figures reproduced from arXiv: 2412.16462 by the authors.

**Figure 2.** Illustration of the Stein gradient (red arrows) and pseudopotential (contours) for [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

**Figure 3.** Error distribution (a, Bhatacharyya distance between the posterior and the true distribution) and sparsity (b, [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Figures from the paper (10 more)

**Figure 4.** Figure 4: Converged particles for various bandwidths using a [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

**Figure 5.** Figure 5: Sequence of graphs for 3 randomly selected particles during the iteration process. Initial models are on the [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

**Figure 6.** Figure 6: Evolution of graph distances of the ensemble during SVGD (blue pixels indicate zero and yellow pixels [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

**Figure 7.** Figure 7: Final kernel density estimates (KDE) of the distribution of weights: (a) [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

**Figure 8.** Figure 8: Accuracy (a, W1 distance) and sparsity (b, parameter count) for various priors with and without graph condensation. Note initial parameter count is 1020. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

**Figure 9.** Figure 9: Convergence of accuracy (a) and sparsity (b) for various prior orders [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

**Figure 10.** Figure 10: Convergence of accuracy (a) and sparsity (b) for various ensemble sizes [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

**Figure 11.** Figure 11: Convergence of accuracy (a) and sparsity (b) for various noise levels [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]

**Figure 12.** Figure 12: Pushforward errors as measured by the W1 distance as a function of deformation F11: (a) for various multiplicative noise levels, and (b) comparison of methods for 10% noise level. In all cases, the model was initialized with 1,020 parameters. Note the high-frequency n…

**Figure 13.** Figure 13: Comparison of (a) accuracy and (b) sparsity for fixed and adaptive penalty. [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Physics Augmented Machine Learning Discovery of Composition-Dependent Constitutive Laws for 3D Printed Digital Materials
cs.LG 2025-07 conditional novelty 6.0 of 10

A physics-augmented neural network learns composition-dependent, rate-dependent stress-strain behavior of PolyJet digital materials, interpolating well to held-out blends but extrapolating poorly.

Reference graph

Works this paper leans on

36 extracted references · 22 canonical work pages · cited by 1 Pith paper

[7]

Improv- ing the performance of stein variational inference through extreme sparsification of physically-constrained neural network models

Govinda Anantha Padmanabha, Jan Niklas Fuhg, Cosmin Safta, Reese E Jones, and Nikolaos Bouklas. Improv- ing the performance of stein variational inference through extreme sparsification of physically-constrained neural network models. Computer Methods in Applied Mechanics and Engineering, 432:117359, 2024

work page 2024
[1]

Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics

Jan Niklas Fuhg, Reese Edward Jones, and Nikolaos Bouklas. Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics. Computer Methods in Applied Mechanics and Engineering, 426:116973, 2024

work page 2024
[2]

Handbook of uncertainty quantification , volume 6

Roger Ghanem, David Higdon, Houman Owhadi, et al. Handbook of uncertainty quantification , volume 6. Springer New York, 2017

work page 2017
[3]

Large sample properties of simulations using latin hypercube sampling

Michael Stein. Large sample properties of simulations using latin hypercube sampling. Technometrics, 29(2):143–151, 1987

work page 1987
[4]

Stein variational gradient descent: A general purpose bayesian inference algorithm

Qiang Liu and Dilin Wang. Stein variational gradient descent: A general purpose bayesian inference algorithm. Advances in neural information processing systems, 29, 2016

2016
[5]

Projected stein variational newton: A fast and scalable bayesian inference method in high dimensions

Peng Chen, Keyi Wu, Joshua Chen, Tom O’Leary-Roseberry, and Omar Ghattas. Projected stein variational newton: A fast and scalable bayesian inference method in high dimensions. Advances in Neural Information Processing Systems, 32, 2019. 16 Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural NetworksA PREPRINT

work page 2019
[6]

A stein varia- tional newton method

Gianluca Detommaso, Tiangang Cui, Youssef Marzouk, Alessio Spantini, and Robert Scheichl. A stein varia- tional newton method. Advances in Neural Information Processing Systems, 31, 2018

work page 2018
[8]

Learning sparse neural networks through l 0 regular- ization

Christos Louizos, Max Welling, and Diederik P Kingma. Learning sparse neural networks through l 0 regular- ization. arXiv preprint arXiv:1712.01312, 2017

arXiv 2017

Show all 36 references

[9]

Stein variational gradient descent as gradient flow

Qiang Liu. Stein variational gradient descent as gradient flow. Advances in neural information processing systems, 30, 2017

2017
[10]

A stochastic stein variational newton method

Alex Leviyev, Joshua Chen, Yifei Wang, Omar Ghattas, and Aaron Zimmerman. A stochastic stein variational newton method. arXiv preprint arXiv:2204.09039, 2022

2022 arXiv
[11]

Pierre, Kevin Linka, and Ellen Kuhl

Jeremy A McCulloch, Skyler R St. Pierre, Kevin Linka, and Ellen Kuhl. On sparse regression, lp-regularization, and automated model discovery. International Journal for Numerical Methods in Engineering , 125(14):e7481, 2024

2024
[12]

Discovering governing equations from data by sparse identification of nonlinear dynamical systems

Steven L Brunton, Joshua L Proctor, and J Nathan Kutz. Discovering governing equations from data by sparse identification of nonlinear dynamical systems. Proceedings of the national academy of sciences, 113(15):3932– 3937, 2016

2016
[13]

Regression shrinkage and selection via the lasso

Robert Tibshirani. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society Series B: Statistical Methodology, 58(1):267–288, 1996

1996
[14]

Bayesian compressive sensing.IEEE Transactions on signal processing, 56(6):2346–2356, 2008

Shihao Ji, Ya Xue, and Lawrence Carin. Bayesian compressive sensing.IEEE Transactions on signal processing, 56(6):2346–2356, 2008

2008
[15]

Bayesian compressive sensing using laplace priors

S Derin Babacan, Rafael Molina, and Aggelos K Katsaggelos. Bayesian compressive sensing using laplace priors. IEEE Transactions on image processing, 19(1):53–63, 2009

2009
[16]

Bayesian compressive sensing via belief propagation

Dror Baron, Shriram Sarvotham, and Richard G Baraniuk. Bayesian compressive sensing via belief propagation. IEEE Transactions on Signal Processing, 58(1):269–280, 2009

2009
[17]

Active sets, nonsmoothness, and sensitivity

Adrian S Lewis. Active sets, nonsmoothness, and sensitivity. SIAM Journal on Optimization , 13(3):702–725, 2002

2002
[18]

An online active set strategy to overcome the limitations of explicit mpc

Hans Joachim Ferreau, Hans Georg Bock, and Moritz Diehl. An online active set strategy to overcome the limitations of explicit mpc. International Journal of Robust and Nonlinear Control: IFAC-Affiliated Journal , 18(8):816–830, 2008

2008
[19]

On the convergence of an active-set method for ℓ1 minimization

Zaiwen Wen, Wotao Yin, Hongchao Zhang, and Donald Goldfarb. On the convergence of an active-set method for ℓ1 minimization. Optimization Methods and Software, 27(6):1127–1146, 2012

2012
[20]

Learning for constrained optimization: Identifying optimal active constraint sets

Sidhant Misra, Line Roald, and Yeesian Ng. Learning for constrained optimization: Identifying optimal active constraint sets. INFORMS Journal on Computing, 34(1):463–480, 2022

2022
[21]

Introduction to markov chain monte carlo

Charles J Geyer. Introduction to markov chain monte carlo. Handbook of markov chain monte carlo , 20116022(45):22, 2011

2011
[22]

B. W. Silverman. Density Estimation for Statistics and Data Analysis. Chapman and Hall, 1986

1986
[23]

A review on data-driven constitutive laws for solids

Jan N Fuhg, Govinda Anantha Padmanabha, Nikolaos Bouklas, Bahador Bahmani, WaiChing Sun, Nikolaos N Vlassis, Moritz Flaschel, Pietro Carrara, and Laura De Lorenzis. A review on data-driven constitutive laws for solids. Archives of Computational Methods in Engineering, pages 1–43, 2024

2024
[24]

Convexity conditions and existence theorems in nonlinear elasticity.Archive for rational mechanics and Analysis, 63:337–403, 1976

John M Ball. Convexity conditions and existence theorems in nonlinear elasticity.Archive for rational mechanics and Analysis, 63:337–403, 1976

1976
[25]

Data-driven tissue mechanics with polyconvex neural ordinary differential equations

Vahidullah Tac, Francisco Sahli Costabal, and Adrian B Tepole. Data-driven tissue mechanics with polyconvex neural ordinary differential equations. Computer Methods in Applied Mechanics and Engineering, 398:115248, 2022

2022
[26]

Polyconvex neural networks for hyperelastic constitutive models: A rectifi- cation approach

Peiyi Chen and Johann Guilleminot. Polyconvex neural networks for hyperelastic constitutive models: A rectifi- cation approach. Mechanics Research Communications, 125:103993, 2022

2022
[27]

A mechanics-informed artificial neural network approach in data- driven constitutive modeling

Faisal As’ad, Philip Avery, and Charbel Farhat. A mechanics-informed artificial neural network approach in data- driven constitutive modeling. International Journal for Numerical Methods in Engineering, 123(12):2738–2759, 2022

2022
[28]

Learning constitutive relations using symmetric positive definite neural networks

Kailai Xu, Daniel Z Huang, and Eric Darve. Learning constitutive relations using symmetric positive definite neural networks. Journal of Computational Physics, 428:110072, 2021. 17 Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural NetworksA PREPRINT

2021
[29]

Polyconvex anisotropic hyperelasticity with neural networks

Dominik K Klein, Mauricio Fern ´andez, Robert J Martin, Patrizio Neff, and Oliver Weeger. Polyconvex anisotropic hyperelasticity with neural networks. Journal of the Mechanics and Physics of Solids , 159:104703, 2022

2022
[30]

Parametrized polyconvex hyperelasticity with physics-augmented neural networks

Dominik K Klein, Fabian J Roth, Iman Valizadeh, and Oliver Weeger. Parametrized polyconvex hyperelasticity with physics-augmented neural networks. Data-Centric Engineering, 4:e25, 2023

2023
[31]

Neural network-based multiscale modeling of finite strain magneto-elasticity with relaxed convexity criteria

Karl A Kalina, Philipp Gebhart, J ¨org Brummund, Lennart Linden, WaiChing Sun, and Markus K ¨astner. Neural network-based multiscale modeling of finite strain magneto-elasticity with relaxed convexity criteria. Computer Methods in Applied Mechanics and Engineering, 421:116739, 2024

2024
[32]

Learning hyperelastic anisotropy from data via a tensor basis neural network

Jan N Fuhg, Nikolaos Bouklas, and Reese E Jones. Learning hyperelastic anisotropy from data via a tensor basis neural network. Journal of the Mechanics and Physics of Solids, 168:105022, 2022

2022
[33]

Machine- learning convex and texture-dependent macroscopic yield from crystal plasticity simulations

Jan N Fuhg, Lloyd van Wees, Mark Obstalecki, Paul Shade, Nikolaos Bouklas, and Matthew Kasemer. Machine- learning convex and texture-dependent macroscopic yield from crystal plasticity simulations. Materialia, 23:101446, 2022

2022
[34]

Input convex neural networks

Brandon Amos, Lei Xu, and J Zico Kolter. Input convex neural networks. In International Conference on Machine Learning, pages 146–155. PMLR, 2017

2017
[35]

Pytorch: An imperative style, high-performance deep learning library

Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019

2019
[36]

Robust estimation of a location parameter

Peter J Huber. Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution, pages 492–518. Springer, 1992. 18

1992

Pith tools

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

[1] [7]

Improv- ing the performance of stein variational inference through extreme sparsification of physically-constrained neural network models

Govinda Anantha Padmanabha, Jan Niklas Fuhg, Cosmin Safta, Reese E Jones, and Nikolaos Bouklas. Improv- ing the performance of stein variational inference through extreme sparsification of physically-constrained neural network models. Computer Methods in Applied Mechanics and Engineering, 432:117359, 2024

work page 2024

[2] [1]

Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics

Jan Niklas Fuhg, Reese Edward Jones, and Nikolaos Bouklas. Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics. Computer Methods in Applied Mechanics and Engineering, 426:116973, 2024

work page 2024

[3] [2]

Handbook of uncertainty quantification , volume 6

Roger Ghanem, David Higdon, Houman Owhadi, et al. Handbook of uncertainty quantification , volume 6. Springer New York, 2017

work page 2017

[4] [3]

Large sample properties of simulations using latin hypercube sampling

Michael Stein. Large sample properties of simulations using latin hypercube sampling. Technometrics, 29(2):143–151, 1987

work page 1987

[5] [4]

Stein variational gradient descent: A general purpose bayesian inference algorithm

Qiang Liu and Dilin Wang. Stein variational gradient descent: A general purpose bayesian inference algorithm. Advances in neural information processing systems, 29, 2016

2016

[6] [5]

Projected stein variational newton: A fast and scalable bayesian inference method in high dimensions

Peng Chen, Keyi Wu, Joshua Chen, Tom O’Leary-Roseberry, and Omar Ghattas. Projected stein variational newton: A fast and scalable bayesian inference method in high dimensions. Advances in Neural Information Processing Systems, 32, 2019. 16 Condensed Stein Variational Gradient Descent for Uncertainty Quantification of Neural NetworksA PREPRINT

work page 2019

[7] [6]

A stein varia- tional newton method

Gianluca Detommaso, Tiangang Cui, Youssef Marzouk, Alessio Spantini, and Robert Scheichl. A stein varia- tional newton method. Advances in Neural Information Processing Systems, 31, 2018

work page 2018

[8] [8]

Learning sparse neural networks through l 0 regular- ization

Christos Louizos, Max Welling, and Diederik P Kingma. Learning sparse neural networks through l 0 regular- ization. arXiv preprint arXiv:1712.01312, 2017

arXiv 2017