Pith. sign in

REVIEW 1 major objections 5 minor 31 references

Efficient and Robust Carath\'{e}odory-Steinitz Pruning of Positive Discrete Measures

T0 review · 1 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper establishes that Carathéodory–Steinitz pruning can be run as a streaming algorithm with O(N^2) storage and O((M−N)N^2 + N^3) total work, and that the pruning map is locally Lipschitz with respect to total variation distance.

desk verdict Solid algorithmic contribution for Carathéodory–Steinitz pruning, but the stability theorem is proven with an exponentially small radius, so the numerical experiments do not actually probe it in the large-M regime. read the letter →

arxiv 2510.14916 v2 pith:Q7B3OPE3 submitted 2025-10-16 math.NA cs.NA

Filip Bělík , Jesse Chan , Akil Narayan This is my paper · ORCID
classification math.NAcs.NA MSC 65D3265F2565Y20
keywords Carathéodory–SteinitzpruningpositivequadraturerulesmeasurecompressionGivensrotationsstreamingalgorithmtotalvariationstabilitymomentpreservationTchakalofftheorem
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's central claim is that Carathéodory–Steinitz pruning—the standard constructive route from an M-point positive quadrature rule to an N-point one that preserves moments—can be made streaming: only a window of N+k rows of the moment matrix is stored, and Givens rotations update the QR factorization of that window in O(N^2) work per pruned node. The total cost is O((M−N)N^2) + O(N^3), linear in the original rule size M, with O(N^2) storage independent of M. On top of the efficiency gain, the paper proves that the pruning map taking an input measure to its pruned output is locally Lipschitz with respect to total variation distance, provided the moment matrix is a Chebyshev system, the input ordering is fixed, and the sign-selection rule has a unique minimizer at every step. The practical upshot is that quadrature rules with billions of nodes can be compressed to a few dozen nodes with modest memory, and the compression is stable under small perturbations of weights or appended nodes.

What carries the argument

The central object is the Vandermonde-like moment matrix V ∈ R^{M×N} whose columns are the N basis functions evaluated at the M nodes. CSP repeatedly finds a nonzero vector in the cokernel of V^T, moves the weight vector along that direction until one weight hits zero, and deletes that node. SCSP restricts the working set to N+k rows at a time, and GSCSP maintains a dense (N+k)×N QR factorization, with Givens-rotation downdates and updates to replace one row per iteration; the cokernel vector is a trailing column of Q, available in O(N^2) work per step.

What would settle it

For the complexity claim, time one full GSCSP run for fixed M and N = 8, 16, 32, 64 and fit the per-iteration cost; if it grows faster than O(N^2), the advertised per-step complexity is wrong. For the stability claim, run SCSP on a Chebyshev-system node set and perturb the weights by total-variation amounts of 10^-8, 10^-6, and 10^-4; if the output distance does not shrink proportionally, the Lipschitz conclusion of Theorem 4.1 is violated.

Watch

Extended reading notes

Core claim

The core discovery is the GSCSP algorithm: a streaming Carathéodory–Steinitz pruning in which the cokernel vector needed at each step is extracted from a maintained QR factorization of a small (N+1)×N window of the Vandermonde-like matrix, and the factorization is updated after each deletion and insertion using Givens rotations. This reduces per-iteration complexity from O(MN^2) to O(N^2) and storage from O(MN) to O(N^2). The paper also proves Theorem 4.1: under a Chebyshev-system assumption on the moment matrix, a fixed node ordering, and a unique minimizer of the sign-selection rule, the SCSP (hence GSCSP) output is locally Lipschitz with respect to total variation distance, so small admis

Load-bearing premise

For the stability half of the claim, the load-bearing premise is that every N-point subset of the nodes gives a nonsingular moment matrix (a Chebyshev system), together with a fixed node ordering and a unique argmin in the sign-selection step; if any N-by-N submatrix is singular, the cokernel vector is non-unique and the pruning path can jump under arbitrarily small perturbations.

Editorial extensions

If this is right

  • A positive M-point rule with M ≫ N can be compressed to N points while exactly preserving N moments using storage that depends only on N, not M.
  • Total runtime is O((M−N)N^2 + N^3), i.e., linear in M, making billion-point compression feasible with modest memory.
  • Small total-variation perturbations that preserve the node ordering—including perturbations that add new nodes with small weights—change the pruned rule by at most a constant times the perturbation size (Theorem 4.1).
  • The same pruning can be applied to non-polynomial spaces V, as long as the moment conditions are well-defined.
  • The algorithm generates positive, moment-preserving quadrature on cut-cell geometries, directly usable in discontinuous Galerkin discretizations.

Reading between the lines

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

  • Because the stability theorem is tied to a fixed ordering of the input nodes, a natural extension is a permutation-invariant or ordering-adaptive variant; the paper gives no such guarantee, and its own analysis shows reordering can change the output.
  • The Chebyshev-system hypothesis fails for structured multivariate node sets (e.g., nodes on a line for monomial bases); a pivoting strategy or random shuffling of inputs might restore uniqueness generically, at the cost of a different stability statement.
  • The Lipschitz constant in the proof grows geometrically through the sequence C_j, so the worst-case bound may be conservative; measuring actual output variation across perturbation scales would show whether practical stability is far better than the theorem guarantees.
  • The O(N^2) storage opens the door to pruning in high-dimensional settings where N is large but M is astronomical, provided the moment matrix can be formed row-by-row; this is exactly the streaming regime the algorithm targets.
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

1 major / 5 minor

Summary. The paper develops a streaming implementation of Carathéodory–Steinitz pruning for compressing a positive M-node quadrature measure to an N-node measure preserving N moments in a finite-dimensional space V. The SCSP algorithm (Alg. 2) retains only N+k rows of the Vandermonde-like matrix at a time, and the GSCSP variant (Alg. 3) uses Givens row downdates/updates so that each iteration costs O(N^2), giving total O((M-N)N^2+N^3) work and O(N^2) storage. Theorem 4.1 asserts local Lipschitz continuity in a normalized total-variation metric under admissible perturbations, assuming a Chebyshev-system condition, fixed ordering, k=1, and unique selection at every iteration. The numerical section compares GSCSP with NNLS and LP, demonstrates large-scale pruning, and includes a cut-cell discontinuous Galerkin application.

Significance. The algorithmic contribution is solid and practically relevant: for the M≫N regime, GSCSP removes the dominant storage bottleneck of naive CSP, and the Givens downdate/update mechanism is an appropriate and well-executed device. The paper provides clear pseudocode, a complexity analysis, a detailed proof of Theorem 4.1 in Appendix A, and reproducible open-source code in CaratheodoryPruning.jl. The stability theorem is a step toward understanding perturbation behavior of pruning algorithms, and the experiments usefully contrast GSCSP, NNLS, and LP. However, as detailed below, the quantitative stability guarantee is currently far too weak to justify the practical 'robust' claim in the advertised large-M regime, and the numerical stability experiments are not actually covered by Theorem 4.1. With an honest reframing of the stability claims, the paper would be acceptable.

major comments (1)
  1. [Section 4.2 / Appendix A, Eq. (21) and the definition of δ2] The stability theorem is stated as a quantitative Lipschitz bound, but its constants are exponential in M-N and the admissible radius is astronomically small in the paper's own target regime. From C_j=(1+N_j)C_{j-1}+1 and N_j≥1, one gets C_{M-N}≥2^{M-N+1}-1; in the Figure 5 example M=10^4, N=113, so C_{M-N}≳2^{9888}≈10^{2977}. Consequently δ2 ≤ (1/(3|µM|)) min_j ϵ_j/C_j is below ~10^{-2977} even if every ϵ_j were of order 1, and the Lipschitz constant C in the theorem contains the same exponential factor. The perturbations tested in Figure 5 begin at d_TV=10^{-10}, so every plotted point lies outside the radius for which the bound is proven. Therefore the statement in Section 5.3 that these experiments provide 'supporting evidence' for the TV-stability theorem is not justified. The theorem may be true, but it currently establishes only qualitative continuity, not practical robustness in
minor comments (5)
  1. [Table 1] The complexity entry for SCSP appears inconsistent with Section 3.1. The table lists per-iteration complexity M N^2 for SCSP, but the text states that fixed-k SCSP computes a cokernel vector of an (N+k)×N matrix at cost O(N^3) per iteration and O((M-N)N^3) total. If the table is per-iteration, the entry should be N^3; if it is cumulative, the CSP and SCSP entries need to be adjusted accordingly.
  2. [Section 5.3] Please temper the language that the numerical stability tests 'support' Theorem 4.1. Since the tested perturbations are many orders of magnitude larger than the proven δ0, the experiments should be described as empirical stability observations in a regime not covered by the theorem, rather than as confirmation of the theorem.
  3. [Algorithm 3] The output line 'k ← M∗{N+1}' appears garbled; presumably it should be 'n ← Q_{*,N+1}' or similar. As printed, it is not readable pseudocode.
  4. [Section 4, intro] The sentence 'This result immediately applies to the GSCSP algorithm' should be qualified: the theorem is proved for the exact SCSP procedure in exact arithmetic. The floating-point Givens updates in GSCSP introduce roundoff, and no roundoff analysis is supplied, so the formal stability guarantee does not automatically transfer to the implemented floating-point algorithm.
  5. [Section 5.2.2] There is a typo: 'Figure Figure 6' should read 'Figure 6'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: GSCSP complexity and SCSP stability are independently derived from explicit algorithms and adjoint perturbation tracking.

full rationale

The paper's central derivation chain is not circular. The SCSP/GSCSP algorithms are explicit, executable procedures (Algorithms 2 and 3), and the claimed O(N^2) storage and O((M-N)N^2 + N^3) complexity are operation counts built from the Givens downdate/update procedure, which is an external standard result [14]; no fitted parameter is later relabeled as a prediction. Theorem 4.1's local-Lipschitz claim is also not assumed in its inputs: Assumption 4.1 and Definition 4.1 only pin down a unique unperturbed pruning path, while Appendix A independently constructs delta0 and C from the optimality gaps epsilon_j, normalized kernel masses N_j, the matrix U, and the weight margin u, then proves by induction that a sufficiently small admissible perturbation leaves the pruning choices unchanged. That is a genuine proof, not a renaming or an imported uniqueness theorem. The self-citations in the manuscript ([5], [21], [23], [24]) are used for applications and context (entropy-stable DG, reduced-order modeling, cut-cell quadrature construction), not as evidence for the new stability bound or complexity estimate. The numerical sections compare GSCSP against external NNLS and LP implementations; the stability experiments are supporting examples rather than the proof. The most serious weakness is that Theorem 4.1's guaranteed perturbation radius shrinks exponentially with M-N (e.g., roughly 2^{-(M-N)}), and the Figure 5 experiments at 10^{-10}-10^{-2} may lie outside that radius. That is a vacuity/strength concern for the 'robust' claim, not circular reasoning. Appendix D also sketches some omitted NNLS proof details, but that result is peripheral and no self-citation is used to fill the gap. No circular step meeting the required evidence standard was found.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced. The paper uses standard Givens rotations and QR factorizations; the only novel objects are algorithmic procedures, not postulated entities.

free parameters (2)
  • k (streaming buffer size offset) = 1
    Algorithm 2 takes a positive integer k; Assumption 4.1 fixes k=1 and all experiments use k=1. Complexity/storage bounds treat k as a constant, so the O(N^2) claims depend on k being small and fixed.
  • τ (Xτ threshold) = arbitrary positive τ (not fitted)
    Definition 4.2 introduces τ to restrict admissible added nodes to where v(x)/||v||_{L1} is bounded; Theorem 4.1 holds for any fixed τ, and δ0/C depend on it. It is a theoretical parameter, not fitted to data.
assumptions (5)
  • standard math Tchakaloff's theorem guarantees a Q≤N-point positive quadrature rule exists for polynomial spaces
    Used in Section 2.1 (Theorem 2.1) to set up the existence target; the paper then develops an algorithm realizing Q=N for finitely supported measures.
  • standard math Carathéodory's conic theorem (Theorem 2.3)
    Basis of the pruning construction: if a vector is in the conic hull of >N points, a subset of ≤N points still contains it in its conic hull.
  • domain assumption V is a Chebyshev system with respect to µM (every N×N submatrix of V is nonsingular)
    Definition 4.1; required for uniqueness of cokernel vectors and for the TV-stability proof of Theorem 4.1.
  • domain assumption Assumption 4.1: k=1, SigSelect (9), unique argmin at each iteration, fixed ordering Σ of original support
    Conditions that make the SCSP output a deterministic function of (V,µM,Σ); Theorem 4.1's Lipschitz statement is conditional on these.
  • domain assumption The initial measure µM is finitely supported or well approximated by a finite measure
    Section 2.2 states the problem setting; the algorithm operates on the discrete measure µM and its moments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient and Robust Carath\'{e}odory-Steinitz Pruning of Positive Discrete Measures." pith.science (2026). https://pith.science/paper/Q7B3OPE3

@misc{pith2026251014916,
  author       = {Pith},
  title        = {Pith review of: Efficient and Robust Carath\'eodory-Steinitz Pruning of Positive Discrete Measures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q7B3OPE3}},
  note         = {Machine review of arXiv:2510.14916}
}
read the original abstract

In many applications, one seeks to approximate integration against a positive measure of interest by a positive discrete measure: a numerical quadrature rule with positive weights. One common desired discretization property is moment preservation over a finite dimensional function space, e.g., bounded-degree polynomials. Carath\'{e}odory's theorem asserts that if there is any finitely supported quadrature rule with more nodes than the dimension of the given function space, one can form a smaller (and hence more efficient) positive, nested, quadrature rule that preserves the moments of the original rule. We describe an efficient streaming procedure for Carath\'{e}odory-Steinitz pruning, a numerical procedure that implements Carath\'{e}odory's theorem for this measure compression. The new algorithm makes use of Givens rotations and on-demand storage of arrays to successfully prune very large rules whose storage complexity only depends on the dimension of the function space. This approach improves on a naive implementation of Carath\'{e}odory-Steinitz pruning whose runtime and storage complexity are quadratic and linear, respectively, in the size of the original measure. We additionally prove mathematical stability properties of our method with respect to a set of admissible, total-variation perturbations of the original measure. Our method is compared to two alternative approaches with larger storage requirements: non-negative least squares and linear programming, and we demonstrate comparable runtimes, with improved stability and storage robustness. Finally, we demonstrate practical usage of this algorithm to generate quadrature for discontinuous Galerkin finite element simulations on cut-cell meshes.

Figures

Figures reproduced from arXiv: 2510.14916 by the authors.

Figure 1
Figure 1. Visual depiction of the two possible pruning choices for given weights and kernel vector. From left to right: visualizing w, n, w − c+n, and w − c−n. One example of the function SigSelect would be the simple rule, SigSelect = argmin σ∈{+,−} |cσ| =⇒ m = argmin m∈S+∪S− wm |nm| , c = wm nm (9) , which simply chooses + or − based on which choice corresponds to a minimum￾norm perturbation of w. In general, this choice co… view at source ↗
Figure 2
Figure 2. Runtime comparison of GSCSP pruning procedure to the LP and NNLS approaches for N = 8 (left) and N = 256 (right) varying M. Each scatter point refers to a mean over 20 trials. and NNLS algorithms are simply infeasible to use due to computational storage re￾quirements. With M = 109 and N ≈ 70, it would take ≈560GB of memory to store the dense Vandermonde matrix in double precision. Let x = (x (1), . . . , x(d) ) ∈ R … view at source ↗
Figure 3
Figure 3. Examples of GSCSP-pruned quadrature rules on various 2D shapes. Left, middle, and right: examples X1, X2, and X3, respectively. shown in [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Example X4 of a GSCSP-pruned quadrature rule on a 3D domain. 5.3. Stability of pruned quadrature rules. We provide empirical experiments to complement the stability theory provided in Section 4. In particular, we inves￾tigate dTV(νM, νeM) when µeM is a small perturbati…
Figure 5
Figure 5. Figure 5: TV errors after various TV perturbations to a discrete measure µM. distances are applied to µM to form µeM. We then compute νeCSP, νeLP, and νeNNLS, by applying the GSCSP, LP, and NNLS algorithms to µeM respectively. Finally, we compute dTV(νCSP, νeCSP), dTV(νLP, νeLP)…
Figure 6
Figure 6. Figure 6: Replication of the middle and right subfigures of Fig￾ure 5 with two choices of the LP c vector. x − 1 0 1 y − 1 0 1 GSCSP Pruned Relative Errors Relati v e Err or 10 − 11 10 − 10 10 − 9 10 − 8 10 − 7 10 − 6 10 − 5 10 − 4 10 − 3 10 − 2 10 − 1 100 x − 1 0 1 y − 1 0 1 GS…
Figure 7
Figure 7. Figure 7: Relative errors and point comparison for pruned per￾turbed quadrature rules on 2D circle. Pruning now done by NNLS. 5.4. Application: cut-cell discontinuous Galerkin (DG) methods. We con￾clude with an example of Caratheodory-Steinitz pruning applied to the generation o…
Figure 8
Figure 8. Figure 8: The cut domain for the advection-diffusion problem (left) and pruned quadrature nodes compared with un-pruned quadrature nodes on a single cut cell (right) [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: Snapshots of the degree P = 7 cut-cell DG solution to the advection-diffusion problem at times t ≈ 0.551 (left), t ≈ 1.7755 (center), and t = 3 (right) with color limits of (0, 0.5), (0, 1.5), and (0, 2.25) respectively. 6. Conclusion We have proposed a new, computatio…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 9 canonical work pages

  1. [1]

    The proof of Tchakaloff’s Theorem

    C. Bayer and J. Teichmann. “The proof of Tchakaloff’s Theorem”. en. In: Proceedings of the American Mathematical Society 134.10 (2006), pp. 3035–

    2006

  2. [2]

    Non-intrusive uncertainty quantification using reduced cubature rules

    L. M. M. van den Bos, B. Koren, and R. P. Dwight. “Non-intrusive uncertainty quantification using reduced cubature rules”. In: Journal of Computational Physics 332 (2017), pp. 418–445. issn: 0021-9991. doi: 10 . 1016 / j . jcp . 2016.12.011

    2017

  3. [3]

    Generating Nested Quadrature Rules with Positive Weights based on Arbitrary Sam- ple Sets

    L. van den Bos, B. Sanderse, W. Bierbooms, and G. van Bussel. “Generating Nested Quadrature Rules with Positive Weights based on Arbitrary Sam- ple Sets”. In: SIAM/ASA Journal on Uncertainty Quantification 8.1 (2020), pp. 139–169. doi: 10.1137/18M1213373

    work page doi:10.1137/18m1213373 2020

  4. [4]

    A fast non-negativity-constrained least squares algo- rithm

    R. Bro and S. De Jong. “A fast non-negativity-constrained least squares algo- rithm”. In: Journal of Chemometrics: A Journal of the Chemometrics Society 11.5 (1997), pp. 393–401

    1997

  5. [5]

    Skew-symmetric entropy stable modal discontinuous Galerkin for- mulations

    J. Chan. “Skew-symmetric entropy stable modal discontinuous Galerkin for- mulations”. In: Journal of Scientific Computing 81.1 (2019), pp. 459–485

    2019

  6. [6]

    A duality proof of Tchakaloff’s theorem

    R. E. Curto and L. A. Fialkow. “A duality proof of Tchakaloff’s theorem”. In: Journal of Mathematical Analysis and Applications 269.2 (2002), pp. 519–

    2002

  7. [7]

    A construction of nonnegative approximate quadratures

    P. J. Davis. “A construction of nonnegative approximate quadratures”. In: Mathematics of Computation 21.100 (1967), pp. 578–582. issn: 0025-5718, 1088-6842. doi: 10.1090/S0025-5718-1967-0222534-4

    work page doi:10.1090/s0025-5718-1967-0222534-4 1967

  8. [8]

    Carath´ eodory bounds for integer cones

    F. Eisenbrand and G. Shmonin. “Carath´ eodory bounds for integer cones”. In: Operations Research Letters 34.5 (2006), pp. 564–568. issn: 0167-6377. doi: https://doi.org/10.1016/j.orl.2005.09.008

    doi:10.1016/j.orl.2005.09.008 2006

Show all 31 references
  1. [9]

    The BR1 scheme is stable for the compressible Navier–Stokes equations

    G. J. Gassner, A. R. Winters, F. J. Hindenlang, and D. A. Kopriva. “The BR1 scheme is stable for the compressible Navier–Stokes equations”. In: Journal of Scientific Computing 77.1 (2018), pp. 154–200

    2018

  2. [10]

    A two-dimensional stabilized discontinuous Galerkin method on curvilinear embedded boundary grids

    A. Giuliani. “A two-dimensional stabilized discontinuous Galerkin method on curvilinear embedded boundary grids”. In: SIAM Journal on Scientific Computing 44.1 (2022), A389–A415

    2022

  3. [11]

    Constructing Positive Interpolatory Cubature Formulas

    J. Glaubitz. “Constructing Positive Interpolatory Cubature Formulas”. In: arXiv:2009.11981 [cs, math] (2020). arXiv: 2009.11981

    2009 arXiv

  4. [12]

    Stable High Order Quadrature Rules for Scattered Data and General Weight Functions

    J. Glaubitz. “Stable High Order Quadrature Rules for Scattered Data and General Weight Functions”. In: SIAM Journal on Numerical Analysis 58.4 (2020), pp. 2144–2164. issn: 0036-1429. doi: 10.1137/19M1257901

    2020 doi

  5. [13]

    Construction and application of provable positive and exact cubature formulas

    J. Glaubitz. “Construction and application of provable positive and exact cubature formulas”. In: IMA Journal of Numerical Analysis 43.3 (2023), pp. 1616–1652. issn: 0272-4979. doi: 10.1093/imanum/drac017

    2023 doi

  6. [14]

    G. H. Golub and C. F. V. Loan. Matrix Computations (Johns Hopkins Studies in Mathematical Sciences) . 3rd. The Johns Hopkins University Press, 1996. isbn: 0-8018-5414-8

    1996

  7. [15]

    J. S. Hesthaven and T. Warburton. Nodal discontinuous Galerkin methods: algorithms, analysis, and applications . Springer

  8. [16]

    Karniadakis and S

    G. Karniadakis and S. Sherwin. Spectral/hp element methods for computa- tional fluid dynamics . Oxford University Press, USA, 2013. 22 REFERENCES

    2013

  9. [17]

    C. L. Lawson and R. J. Hanson. Solving Least Squares Problems . Society for Industrial and Applied Mathematics, 1995. doi: 10.1137/1.9781611971217. eprint: https://epubs.siam.org/doi/pdf/10.1137/1.9781611971217

    1995 doi

  10. [18]

    JuMP 1.0: Recent improvements to a modeling language for mathemati- cal optimization

    M. Lubin, O. Dowson, J. Dias Garcia, J. Huchette, B. Legat, and J. P. Vielma. “JuMP 1.0: Recent improvements to a modeling language for mathemati- cal optimization”. In: Mathematical Programming Computation 15 (2023), pp. 581–589. doi: 10.1007/s12532-023-00239-3

    2023 doi

  11. [19]

    Caratheodory- Tchakaloff Subsampling

    F. Piazzon, A. Sommariva, M. Vianello, and M. Vianello. “Caratheodory- Tchakaloff Subsampling”. In: Dolomites Research Notes on Approximation 10.1 (2017). arxiv 1611.02065 [math.NA], pp. 5–14. issn: 20356803. doi: 10. 14658/pupj-drna-2017-1-2

    2017 arXiv

  12. [20]

    A note on Tchakaloff’s Theorem

    M. Putinar. “A note on Tchakaloff’s Theorem”. In: Proceedings of the Ameri- can Mathematical Society 125.8 (1997), pp. 2409–2414. issn: 0002-9939, 1088-

    1997

  13. [21]

    Entropy stable reduced order modeling of nonlinear conservation laws using discontinuous Galerkin methods

    R. Qu, A. Narayan, and J. Chan. “Entropy stable reduced order modeling of nonlinear conservation laws using discontinuous Galerkin methods”. In:arXiv preprint arXiv:2502.09381 (2025)

    2025

  14. [22]

    Bedingt konvergente Reihen und konvexe Systeme

    E. Steinitz. “Bedingt konvergente Reihen und konvexe Systeme.” de. In: Jour- nal f¨ ur die reine und angewandte Mathematik 1913.143 (1913), pp. 128–176. issn: 1435-5345. doi: 10.1515/crll.1913.143.128

    1913 doi

  15. [23]

    An Entropy Stable High-Order Discontinuous Galerkin Method on Cut Meshes

    C. G. Taylor and J. Chan. “An Entropy Stable High-Order Discontinuous Galerkin Method on Cut Meshes”. In:arXiv preprint arXiv:2412.13002 (2024)

    2024 arXiv

  16. [24]

    An energy stable high-order cut cell discontinuous Galerkin method with state redistribution for wave propa- gation

    C. G. Taylor, L. C. Wilcox, and J. Chan. “An energy stable high-order cut cell discontinuous Galerkin method with state redistribution for wave propa- gation”. In: Journal of Computational Physics 521 (2025), p. 113528

    2025

  17. [25]

    Formules de cubatures m´ ecaniques ` a coefficients non n´ egatifs

    V. Tchakaloff. “Formules de cubatures m´ ecaniques ` a coefficients non n´ egatifs”. In: Bull. Sci. Math 81.2 (1957), pp. 123–134

    1957

  18. [26]

    Caratheodory cubature measures

    M. Tchernychova. “Caratheodory cubature measures”. http://purl.org/dc/dcmitype/Text. University of Oxford, 2016

    2016

  19. [27]

    A general algorithm for nonnegative quadrature formulas

    M. W. Wilson. “A general algorithm for nonnegative quadrature formulas”. In: Mathematics of Computation 23.106 (1969), pp. 253–258. issn: 0025-5718, 1088-6842. doi: 10.1090/S0025-5718-1969-0242374-1

    1969 doi

  20. [28]

    An LP empirical quadrature procedure for re- duced basis treatment of parametrized nonlinear PDEs

    M. Yano and A. T. Patera. “An LP empirical quadrature procedure for re- duced basis treatment of parametrized nonlinear PDEs”. In: Computer Meth- ods in Applied Mechanics and Engineering 344 (2019), pp. 1104–1123. issn: 0045-7825. doi: https://doi.org/10.1016/j.cma.2018.02.028...

    2019 doi

  21. [532]

    doi: 10.1016/S0022-247X(02)00034-3

    issn: 0022-247X. doi: 10.1016/S0022-247X(02)00034-3

    doi

  22. [3040]

    doi: 10.1090/S0002-9939-06-08249-9

    issn: 0002-9939, 1088-6826. doi: 10.1090/S0002-9939-06-08249-9

    doi

  23. [6826]

    doi: 10.1090/S0002-9939-97-03862-8

    doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.