Pith. sign in

REVIEW 3 major objections 5 minor 25 references

Enhancing finite-difference based derivative-free optimization methods with machine learning

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A surrogate-model wrapper around finite-difference derivative-free optimization reduces the worst-case function-evaluation count, and its leading term becomes $O(\epsilon^{-2})$, independent of dimension, when successful surrogate steps…

desk verdict A genuinely useful surrogate-acceleration paper for finite-difference DFO with a solid complexity theorem; the main caveat is that the numerical gains are measured in oracle calls, not wall-clock time, so the training overhead is unaccounted. read the letter →

arxiv 2502.07435 v1 pith:ZBSHFP3C submitted 2025-02-11 math.OC

Timothé Taminiau , Estelle Massart , Geovani Nunes Grapiglia This is my paper · ORCID
classification math.OC MSC 90C5690C3065K05
keywords derivative-freeoptimizationfinite-differencegradientssurrogatemodelsSobolevlearningworst-casecomplexityneuralnetworksradialbasisfunctionsdataprofiles
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

Derivative-free optimization methods that replace gradients by finite differences automatically accumulate a dataset of points, function values, approximate gradients, and step sizes. The paper's proposal is to train a surrogate model on that dataset, including the approximate gradients through Sobolev learning, and then take cheap surrogate-driven gradient steps until they stop decreasing the true objective, at which point the base method resumes. Integrated with a finite-difference method from the literature, the heuristic finds an $\epsilon$-approximate stationary point with at most $O(n\epsilon^{-2})$ function evaluations, and when the average number of successful surrogate steps per outer iteration is at least $n$, the leading term becomes $O(\epsilon^{-2})$, independent of dimension. Each attempted surrogate step costs one function evaluation, which is why a surrogate that is accepted more often shrinks the constant in the worst-case bound. Numerical data profiles on 134 problems from the OPM/CUTEst benchmark show that neural-network and radial-basis surrogates solve more problems within a fixed budget than the base method, especially when approximate gradients are used in training.

What carries the argument

The carrying object is Algorithm 1, the surrogate step: after the base method computes a finite-difference gradient, the datasets $F$ and $G$ are updated and a surrogate $m_\theta$ is trained by minimizing the Sobolev-learning objective (2), which penalizes both mismatches to stored function values and mismatches of $\nabla m_\theta$ to stored finite-difference gradients. The decisive property is that each accepted surrogate step forces a decrease of at least $\epsilon^2/(2\gamma\sigma_{\max})$ in the true objective while costing one function evaluation, so the proof can add these decreases to the base method's own decrease. Proposition 3.1 identifies the geometric content of the Sobolev penalty: if the surrogate interpolates the finite-difference points exactly, the gradient-matching term equals a second-order Taylor remainder, so Sobolev learning with finite-difference gradients is a penalty on the surrogate's curvature.

What would settle it

A decisive experiment would count surrogate-training effort in the budget: on a cheap-to-evaluate problem, track total wall-clock time and total L-BFGS-iteration cost for Algorithm 3 with a neural-network surrogate versus Algorithm 2, both run until $\|\nabla f(x_k)\|_2 \le \epsilon$; if the accelerated method is slower once training is included, the practical efficiency claim fails even though Theorem 2.4 remains correct as an oracle-counting statement. A simpler version would repeat the data profiles with the x-axis measured in units that include surrogate training cost rather than simplex gradients.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a simple wrapper provably improves finite-difference DFO: train a continuously differentiable surrogate on the collected points and finite-difference gradients, then apply gradient descent with Armijo line search to the surrogate until it fails to ensure sufficient decrease of the true objective, and accept each such step as a new iterate. Theorem 2.4 bounds the number of function evaluations by $4\eta(S)(n+1)C_{\max}(f(x_0)-f_{\mathrm{low}})\epsilon^{-2} + \log_2(\sigma_{\max}/\sigma_0)(n+1) + T(\epsilon)$, where $\eta(S) = (1 + S/(2(n+1)))/(1+S)$ and $S$ is the average number of successful surrogate steps per outer iteration. Because $\eta(S)$ decreases with $S$, a surrogate that is accepted more often gives a strictly better oracle-complexity bound, and for $S \ge n$ the leading term becomes $O(\epsilon^{-2})$ with no factor of $n$. The paper also shows numerically, using radial-basis and shallow neural-network surrogates on 134 benchmark problems, that the accelerated method outperforms the base method, and that Sobolev learning, which includes approximate gradients in the training objective, strengthens the improvement.

Load-bearing premise

The practical acceleration rests on the assumption that training or updating the surrogate is negligible compared with one evaluation of $f$: the theorem counts only zeroth-order oracle calls, and the numerical plots count only simplex gradients, so surrogate training, which for neural networks is an L-BFGS solve of a nonconvex problem run up to 1000 iterations per outer step, is not part of the budget. If function evaluations are cheap relative to training, the apparent speedup can vanish.

Editorial extensions

If this is right

  • Each attempted surrogate step consumes one function evaluation, and each accepted step contributes a fixed $\epsilon^2$-scale decrease in the objective, so more reliable surrogates translate directly into a smaller worst-case evaluation count.
  • If the average number of successful surrogate steps per outer iteration reaches $n$, the leading term in the evaluation bound becomes $6C_{\max}(f(x_0)-f_{\mathrm{low}})\epsilon^{-2}$, independent of the dimension; only a logarithmic $n$-term from the unknown Lipschitz constant remains.
  • The heuristic cannot worsen the theoretical worst case: if no surrogate step is ever accepted, the bound reduces to the base method's $O(n\epsilon^{-2})$ with the same constants.
  • Including finite-difference gradients in surrogate training is numerically beneficial: Sobolev learning gives lower surrogate-gain values and better data profiles than training on function values alone.
  • On the 134-problem benchmark, the surrogate-accelerated methods solve a larger proportion of problems within any fixed budget of simplex gradients than the base method, and the gap widens as the budget grows.

Reading between the lines

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

  • My inference: the same wrapper should accelerate other finite-difference DFO methods, since the proof of Theorem 2.4 only needs the base method to supply a per-iteration decrease lower bound and bounded finite-difference stepsizes; any method satisfying those two inequalities inherits the $\eta(S)$ gain.
  • My inference: Proposition 3.1 suggests a design rule the paper does not state explicitly; when interpolation is inexact, the surrogate's curvature along coordinate directions is still the quantity Sobolev learning controls, so choosing models with bounded second derivatives should make the accepted surrogate steps more reliable.
  • My inference: because the complexity theorem and the data profiles count only function evaluations, an apples-to-apples comparison for cheap objectives should include surrogate-training work; the paper's own remark that RBF training is much cheaper than neural-network training implies that a wall-clock benchmark could shift the NN-versus-RBF ranking on problems where evaluations are fast.
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

3 major / 5 minor

Summary. This paper proposes a surrogate-based acceleration heuristic for finite-difference derivative-free optimization (DFO). The heuristic (Algorithm 1) trains a surrogate model using the accumulated dataset of function values and finite-difference gradient approximations (Sobolev learning, equation (2)), and then performs gradient descent with Armijo line search on the surrogate until a step fails to produce sufficient decrease on the true objective, at which point control returns to the base DFO method and the dataset is augmented. The authors integrate this heuristic with a simplified version of the derivative-free quadratic regularization method of Grapiglia (Algorithm 2), yielding Algorithm 3. The main theoretical result is Theorem 2.4, which bounds the number of function evaluations required to reach an epsilon-approximate stationary point by 4*eta(S)*(n+1)*C_max*(f(x0)-f_low)*epsilon^{-2} + log2(sigma_max/sigma0)*(n+1) + T(epsilon), where eta(S) = (1 + S/(2(n+1)))/(1+S) decreases with the average number of successful surrogate steps S; when S >= n, the leading term becomes independent of n. Numerical experiments on 134 CUTEst problems compare NN and RBF surrogates, with and without Sobolev learning, against the base method using data profiles measured in simplex gradients.

Significance. If Theorem 2.4 is correct, the paper provides a clean and honest oracle-complexity certification for a class of ML-accelerated DFO methods: the bound depends on the observed number of successful surrogate steps rather than assuming it, and each successful surrogate step provably costs only one function evaluation. The proof of the main theorem is mostly elementary and the accounting of function evaluations is consistent. Proposition 3.1 gives a neat interpretation of Sobolev training with finite-difference gradients as curvature regularization. The numerical section is thorough in its coverage of model families and ablations, and the reported median surrogate gains (eta approximately 0.3-0.4 with Sobolev learning) support the qualitative prediction of the theory. However, the practical significance of the acceleration depends on the assumption that oracle evaluations dominate the total computational cost, because the NN training cost (L-BFGS up to 1000 iterations per outer loop) is excluded from both the theory and the benchmarks. The paper would be substantially strengthened by a wall-clock-normalized comparison or at least a report of training costs.

major comments (3)
  1. [Section 4.2, Figures 1-4] The data profiles in Figures 1, 2, and 4 count only simplex gradients, and Theorem 2.4 counts only evaluations of f; the cost of training the surrogate by solving (2) with L-BFGS (up to 1000 iterations per call, Section 4.2) is invisible to both the theory and the benchmarks. Algorithm 3 invokes Surrogate at every outer iteration, so if function evaluations are cheap relative to training, the reported gains in simplex gradients may not translate into wall-clock speedups. The paper acknowledges this asymmetry only in the RBF-versus-NN comparison in Section 4.3, not in the accelerated-versus-base comparison that motivates the abstract's claim of 'significant performance improvements.' Please report training time per run, a wall-clock-normalized data profile, or at least the number of L-BFGS iterations and average training time per problem, so the reader can judge whether the motivating expensive-evaluation regime is represented.
  2. [Lemma 2.2] Lemma 2.2 is the foundation of the complexity analysis, but its proof is entirely delegated: 'It follows directly from Lemmas 1-4 in [13].' Since Algorithm 2 is a simplified version of the method in [13] (it sets B_k=0 and uses a particular Armijo backtracking), and since the constants C_f and sigma_max appear in Theorem 2.4, the manuscript should either restate the relevant lemmas from [13] and verify that their hypotheses hold for Algorithm 2, or provide a self-contained proof in an appendix. As written, a reader cannot verify the load-bearing estimates (4) and (5) without consulting [13].
  3. [Section 4.2] The neural network surrogate is trained with L-BFGS starting from He or Glorot random initialization, but the paper does not report the random seeds used, and the data profiles appear to be based on a single run per problem. Because the NN training problem (2) is nonconvex, the surrogate quality and hence the number of successful steps could vary across initializations. Please report the seeds and, ideally, show results over multiple runs with confidence bands, at least for a subset of the test problems, to confirm that the observed improvements are stable.
minor comments (5)
  1. [Section 4.3] The sentence 'To validate the worst-case complexity bounds derived in Section 2' overstates what is done; computing the empirical surrogate gain eta(S(T_max)) illustrates the factor in the bound but does not validate the bound, since C_max depends on the unknown Lipschitz constant L. Please rephrase, for example, 'To illustrate the behavior of the surrogate gain in the complexity bound.'
  2. [Algorithm 1] The initialization 'F + := empty set' uses a plus sign as a superscript that is inconsistent with the later notation F^+; please use a consistent symbol.
  3. [Proof of Theorem 2.4] The notation 'log' in the telescoping line should be 'log2' for consistency with the rest of the paper.
  4. [Theorem 2.4] The expression eta(S(T(epsilon)) is missing a closing parenthesis; it should be eta(S(T(epsilon))).
  5. [Section 4.2] The bullet point about the minimal-norm solution for RBF models is helpful, but it should state whether the least-squares solver uses a QR or SVD-based method to compute that solution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 2.4 is a conditional counting bound, and the only self-citation (Lemma 2.2 from [13]) is an independent prior result about the base method.

full rationale

The central oracle-complexity claim (Theorem 2.4) is derived by counting function evaluations: each finite-difference outer iteration costs (i_k+1)(n+1) evaluations, each accepted surrogate step costs one evaluation of f, and the failed surrogate trial is accounted for by the +T(epsilon) term. The factor eta(S) is defined as a decreasing function of the observed average number of successful surrogate steps S(T(epsilon)); it is measured ex post from the run, not fitted to enforce a target improvement, so the bound is not circular. Lemma 2.2 is imported from Lemmas 1-4 of [13], a prior paper by one of the present authors; although this is a self-citation, [13] is an independent published analysis of the base finite-difference method and does not assume the surrogate acceleration whose complexity is being proved. The numerical comparison counts only zeroth-order oracle calls and excludes NN/RBF training time; the paper itself flags this in Section 4.3, noting that for cheap evaluations the RBF-accelerated method may become substantially faster than the NN-accelerated method. That asymmetry is a benchmarking limitation, not a circular derivation. Accordingly, no load-bearing step reduces to its own inputs.

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

No new physical or mathematical entities are introduced; the surrogate model is a computational artifact. The free parameters are algorithm constants and dataset caps fixed before running the benchmark. The central theoretical result depends on standard smoothness and boundedness assumptions, plus the modeling assumption that surrogate training is not part of the oracle cost.

free parameters (4)
  • gamma (surrogate decrease threshold constant) = 25/2
    Fixed in Section 4.2; appears in Algorithm 1's acceptance rule (3) and in the complexity bound of Theorem 2.4. Chosen by hand, not fitted to the benchmark.
  • lambda (Sobolev regularization weight) = 1e-4
    Regularization weight in training objective (2); set a priori in Section 4.2.
  • Dataset caps N <= 10(n+1), M <= 10 = N cap 10(n+1), M cap 10
    Memory and conditioning limits on F and G in Section 4.2; they affect surrogate quality and the observed numerical gains.
  • Algorithmic constants (sigma0, sigma_min, rho, epsilon) = sigma0=1, sigma_min=1e-2, rho=1e-4, epsilon=1e-5
    Standard fixed hyperparameters of the base method and heuristic; chosen by hand and not fitted to the test set.
assumptions (4)
  • domain assumption f is differentiable and bounded below (Assumption A1)
    Stated in Assumption A1 and used in Lemma 2.1 for finite termination of the surrogate loop and in the complexity theorem.
  • domain assumption nabla f is L-Lipschitz continuous (Assumption A2)
    Stated in Assumption A2 and used in Lemma 2.2 to bound sigma_k and the base-method decrease; the estimate is imported from [13].
  • ad hoc to paper The surrogate model m_theta is continuously differentiable
    Required in Step 1 of Algorithm 1 to run Armijo line search on m_theta, which excludes ReLU activations and is stated in Section 4.2.
  • domain assumption Oracle cost dominates surrogate training cost
    Theorem 2.4 counts only evaluations of f, and the data profiles also count only simplex gradients. NN training via L-BFGS up to 1000 iterations is outside the oracle complexity, which matters when evaluations are cheap.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing finite-difference based derivative-free optimization methods with machine learning." pith.science (2026). https://pith.science/paper/ZBSHFP3C

@misc{pith2026250207435,
  author       = {Pith},
  title        = {Pith review of: Enhancing finite-difference based derivative-free optimization methods with machine learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZBSHFP3C}},
  note         = {Machine review of arXiv:2502.07435}
}
read the original abstract

Derivative-Free Optimization (DFO) involves methods that rely solely on evaluations of the objective function. One of the earliest strategies for designing DFO methods is to adapt first-order methods by replacing gradients with finite-difference approximations. The execution of such methods generates a rich dataset about the objective function, including iterate points, function values, approximate gradients, and successful step sizes. In this work, we propose a simple auxiliary procedure to leverage this dataset and enhance the performance of finite-difference-based DFO methods. Specifically, our procedure trains a surrogate model using the available data and applies the gradient method with Armijo line search to the surrogate until it fails to ensure sufficient decrease in the true objective function, in which case we revert to the original algorithm and improve our surrogate based on the new available information. As a proof of concept, we integrate this procedure with the derivative-free method proposed in (Optim. Lett. 18: 195--213, 2024). Numerical results demonstrate significant performance improvements, particularly when the approximate gradients are also used to train the surrogates.

Figures

Figures reproduced from arXiv: 2502.07435 by the authors.

Figure 1
Figure 1. Data profiles comparing the base method to the surrogate-accelerated methods based on [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Data profiles comparing the base method to the surrogate-accelerated methods based on [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Box plots for the distribution of the surrogate gain over the full test set with a budget of [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Data profiles comparing the base method, the surrogate-accelerated methods based on [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Data profiles for the tolerance [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Data profiles for the tolerance τ = 10−4 and budget of 100 simplex gradients. We compare different choices of RBF models for the accelerated method with Sobolev learning. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 24 canonical work pages

  1. [9]

    The limitation of neural nets for approximation and optimization

    Tommaso Giovannelli, Oumaima Sohab, and Luis Nunes Vicente. The limitation of neural nets for approximation and optimization, 2023. arXiv:2311.12253

    work page Pith review arXiv 2023

  2. [13]

    Grapiglia

    Geovani N. Grapiglia. Worst-case evaluation complexity of a derivative-free quadratic regular- ization method. Optimization Letters, 18:1–19, 2024

    work page 2024

  3. [1]

    Charles Audet and J. E. Dennis. Mesh Adaptive Direct Search Algorithms for Constrained Optimization. SIAM Journal on Optimization , 17(1):188–217, 2006. Publisher: Society for Industrial and Applied Mathematics

    work page 2006

  4. [2]

    Derivative-Free and Blackbox Optimization

    Charles Audet and Warren Hare. Derivative-Free and Blackbox Optimization . Springer Series in Operations Research and Financial Engineering. Springer International Publishing, Cham, 2017

    work page 2017

  5. [3]

    A. S. Berahas, O. Sohab, and L. N. Vicente. Full-low evaluation methods for derivative-free optimization. Optimization Methods and Software , 38(2):386–411, 2023

    work page 2023

  6. [4]

    Abdalla, Azizallah Izady, Mohammad Reza Nikoo, and Ali Al- Maktoumi

    Mingjie Chen, Osman A. Abdalla, Azizallah Izady, Mohammad Reza Nikoo, and Ali Al- Maktoumi. Development and surrogate-based calibration of a CO2 reservoir model. Journal of Hydrology, 586(124798):124798, 2020

    work page 2020

  7. [5]

    Conn, Katya Scheinberg, and Luis N

    Andrew R. Conn, Katya Scheinberg, and Luis N. Vicente. Introduction to Derivative-Free Opti- mization. MOS-SIAM Series on Optimization. Society for Industrial and Applied Mathematics, 2009

    work page 2009

  8. [6]

    Conn, Katya Scheinberg, and Lu ´ ıs N

    Andrew R. Conn, Katya Scheinberg, and Lu ´ ıs N. Vicente. Global Convergence of General Derivative-Free Trust-Region Algorithms to First- and Second-Order Critical Points. SIAM Journal on Optimization , 20(1):387–415, 2009

    work page 2009

Show all 25 references
  1. [7]

    Sobolev Training for Neural Networks

    Wojciech Marian Czarnecki, Simon Osindero, Max Jaderberg, Grzegorz ´Swirszcz, and Razvan Pascanu. Sobolev Training for Neural Networks. page arXiv:1706.04859, 2017

    2017 arXiv

  2. [8]

    Frazier and Jialei Wang

    Peter I. Frazier and Jialei Wang. Bayesian optimization for materials design. arXiv e-prints , 225:arXiv:1506.01349, 2016

    2016 arXiv

  3. [10]

    Understanding the difficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the Thirteenth International Conference on Artificial Intel- ligence and Statistics , pages 249–256. JMLR Workshop and Conference Proceedings, 2010

    2010

  4. [11]

    Nicholas I. M. Gould, Dominique Orban, and Philippe L. Toint. CUTEst: a Constrained and Unconstrained Testing Environment with safe threads for mathematical optimization. Compu- tational Optimization and Applications , 60(3):545–557, 2015. 15

    2015

  5. [12]

    Grapiglia

    Geovani N. Grapiglia. Quadratic regularization methods with finite-difference gradient approx- imations. Computational Optimization and Applications , 85:683–703, 2023

    2023

  6. [14]

    Serge Gratton and Philippe L. Toint. OPM, a collection of Optimization Problems in Matlab. arXiv e-prints , page arXiv:2112.05636, 2021

    2021 arXiv

  7. [15]

    Temperature modeling of creep- feed grinding processes for nickel-based superalloys with variable heat flux distribution

    Adina Grimmert, Florian Pachnek, and Petra Wiederkehr. Temperature modeling of creep- feed grinding processes for nickel-based superalloys with variable heat flux distribution. CIRP Journal of Manufacturing Science and Technology , 41:477–489, 2023

    2023

  8. [16]

    Delving Deep into Rectifiers: Sur- passing Human-Level Performance on ImageNet Classification

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delving Deep into Rectifiers: Sur- passing Human-Level Performance on ImageNet Classification. In 2015 IEEE International Conference on Computer Vision (ICCV) , pages 1026–1034, 2015

    2015

  9. [17]

    Neural Network Accelerated Implicit Fil- tering: Integrating Neural Network Surrogates With Provably Convergent Derivative Free Op- timization Methods

    Brian Irwin, Eldad Haber, Raviv Gal, and Avi Ziv. Neural Network Accelerated Implicit Fil- tering: Integrating Neural Network Surrogates With Provably Convergent Derivative Free Op- timization Methods. In Proceedings of the 40th International Conference on Machine Learning , p...

    2023

  10. [18]

    C. T. Kelley. Implicit Filtering . Software, Environments, and Tools. Society for Industrial and Applied Mathematics, 2011

    2011

  11. [19]

    ChemBO: Bayesian Optimization of Small Organic Molecules with Synthesizable Recommendations

    Ksenia Korovina, Sailun Xu, Kirthevasan Kandasamy, Willie Neiswanger, Barnabas Poczos, Jeff Schneider, and Eric Xing. ChemBO: Bayesian Optimization of Small Organic Molecules with Synthesizable Recommendations. In Proceedings of the Twenty Third International Conference on Art...

    2020

  12. [20]

    Jeffrey Larson, Matt Menickelly, and Stefan M. Wild. Derivative-free optimization methods. Acta Numerica, 28:287–404, 2019

    2019

  13. [21]

    Benchmarking Derivative-Free Optimization Algorithms

    Jorge Mor´ e and Stefan Wild. Benchmarking Derivative-Free Optimization Algorithms. SIAM Journal on Optimization , 20:172–191, 2009

    2009

  14. [22]

    M. J. D. Powell. The NEWUOA software for unconstrained optimization without derivatives. In G. Di Pillo and M. Roma, editors, Large-Scale Nonlinear Optimization, pages 255–297. Springer US, Boston, MA, 2006

    2006

  15. [23]

    On the numerical performance of finite-difference-based methods for derivative-free optimization

    Hao-Jun Michael Shi, Melody Qiming Xuan, Figen Oztoprak, and Jorge Nocedal. On the numerical performance of finite-difference-based methods for derivative-free optimization. Op- timization Methods and Software , 38(2):289–311, 2023

    2023

  16. [24]

    L. N. Vicente. Worst case complexity of direct search. EURO Journal on Computational Optimization, 1(1):143–153, 2013

    2013

  17. [25]

    Accelerating Discovery of Polyimides with Intrinsic Microporosity for Membrane-Based Gas Separation: Synergizing Physics-Informed Performance Metrics and Active Learning

    Mao Wang and Jianwen Jiang. Accelerating Discovery of Polyimides with Intrinsic Microporosity for Membrane-Based Gas Separation: Synergizing Physics-Informed Performance Metrics and Active Learning. Advanced Functional Materials, 34(23):2314683, 2024. 16 A Supplementary numeri...

    2024

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.