Gromov-Wasserstein Quantization and Clustering: Structure, Rates, and Algorithms

Florian Beier; Stephan Eckstein

Gromov-Wasserstein Quantization and Clustering: Structure, Rates, and Algorithms

T0 review · 0 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Gromov-Wasserstein quantization approximates a space and its geometry together, and in Euclidean settings achieves the same sharp $n^{-1/d}$ rate as classical Wasserstein quantization.

desk verdict A genuinely new and mostly airtight extension of quantization to Gromov–Wasserstein geometry; the rate results are sharp under standard full-dimensionality assumptions that the paper states clearly. read the letter →

arxiv 2608.11016 v1 pith:2GS7NXPI submitted 2026-08-11 math.OC cs.CGcs.LGmath.PR

Florian Beier , Stephan Eckstein This is my paper · ORCID

classification math.OCcs.CGcs.LGmath.PR MSC 49Q2290C2662H30

keywords Gromov-WassersteinquantizationgaugedmeasurespacesratesconditionalgradientclusteringMongemapsneuralnetworkpruningoptimaltransport

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 establishes a Gromov-Wasserstein (GW) analogue of classical quantization: rather than approximating a measure by $n$ points in a fixed space, it approximates a gauged measure space—a measure together with a pairwise gauge such as a distance or an inner product—by an $n$-point space measured with the GW distance, which compares gauges rather than points. The main structural claim is that a best $n$-point approximation always exists and has the same backbone as Wasserstein quantization: the matrix of gauge values between quantized points is the block-mean of the original gauge over cluster cells, and the optimal coupling is concentrated on Voronoi cells of cost functions induced by that gauge. The main quantitative claim is that in Euclidean spaces with the distance, squared-distance, or scalar-product gauge, the GW quantization value is bounded above and below by constant multiples of the Wasserstein quantization value, yielding exactly the sharp $n^{-1/d}$ rate. Along the way the paper lifts the $k$-means centroid iteration to this setting as a conditional-gradient algorithm with monotone decrease and stationary cluster points, so the theoretical structure is matched by a practical method.

What carries the argument

The load-bearing object is the gauged measure space $\mathcal X=(X,g,\xi)$ and the GW quantization value $q^{GW}_n(\mathcal X)=\inf_{\mathcal Y\in GM_n}GW_2(\mathcal X,\mathcal Y)$, where $GM_n$ is the set of $n$-point spaces with a symmetric gauge matrix and arbitrary weights. The identity that carries the argument is the block-mean gauge formula: for any fixed coupling, the best gauge matrix is $\hat G_{i,i'}=\frac{1}{\pi_i(X)\pi_{i'}(X)}\int\int g\,d\pi_i\,d\pi_{i'}$, exactly the $L^2$ mean analogous to the centroid of a $k$-means cluster. Orthogonal to that, the Voronoi characterization says the optimal coupling slices concentrate on the cells of the cost functions $\hat c_i$; together the two form the fixed-point system that the algorithm alternates between. The rates rest on two-sided inequalities comparing the GW value to $q^W_n(\xi)$, with the lower bounds coming from a within-cell variance identity and the upper bounds from Lipschitz continuity of the gauge as a function of the metric.

What would settle it

Take a full-dimensional absolutely continuous measure on $[0,1]^d$ for $d=1,2,3$ with one of the three Euclidean gauges, run the paper's algorithm for increasing $n$, and fit the slope of $\log q^{GW}_n$ against $\log n$; the claim predicts slopes within about $0.02$ of $-1/d$, and a systematic deviation (or a full-rank absolutely continuous sequence with $q^{GW}_n/q^W_n\to\infty$) would refute the two-sided rate bounds.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that GW quantization inherits the complete structure of Wasserstein quantization. Theorem 3.4 proves existence and characterizes every optimal quantizer by two self-consistent conditions: the gauge matrix entry $\hat G_{i,i'}$ equals the $L^2$ mean of $g$ under $\hat\pi_i\otimes\hat\pi_{i'}$, and each $\hat\pi_i$ is concentrated on the Voronoi cell of the induced cost $\hat c_i(x)=\sum_{i'}\int_X(g(x,x')-\hat G_{i,i'})^2\,d\hat\pi_{i'}(x')$. Section 4 proves two-sided bounds $c\,q^W_n(\xi)\le q^{GW}_n(\mathcal X)\le C\,q^W_n(\xi)$ for the three Euclidean gauges, with $c,C$ depending on the spectrum of the covariance or second-moment matrix; by Zador's theorem these give the sharp rate $q^{GW}_n(\mathcal X)\asymp n^{-1/d}$. For the scalar-product gauge the problem reduces to a partition problem in which the gauge value between clusters is the scalar product of cluster means, an optimal Monge map exists, and in one dimension the value is exactly $q^W_n(T_\#\xi)\sqrt{2M_2(T_\#\xi)-q^W_n(T_\#\xi)^2}$.

Load-bearing premise

The sharp $n^{-1/d}$ lower bounds require the measure's covariance (or second-moment) matrix to have a positive smallest eigenvalue, so they degenerate for measures supported on lower-dimensional affine subspaces; the rate results also inherit Zador's conditions of an absolutely continuous component and a finite $(2+\delta)$-moment.

Editorial extensions

If this is right

Every gauged measure space admits an $n$-point GW quantizer, and the optimal quantizer always has the block-mean gauge plus Voronoi-cell coupling form, so GW quantization is a well-posed clustering problem rather than just an optimization heuristic.
For the Euclidean distance, squared distance, and scalar-product gauges, $q^{GW}_n(\mathcal X)$ is sandwiched between constant multiples of $q^W_n(\xi)$, so compressing a space to $n$ representative points retains its geometry at the same sharp $n^{-1/d}$ rate as classical quantization.
In the scalar-product case an optimal solution is induced by a partition, with the quantized gauge given by scalar products of cluster means, and the optimal transport plan is a Monge map, so no mass splitting is needed.
In one dimension, the scalar-product GW quantization value is exactly $q^W_n(T_\#\xi)\sqrt{2M_2(T_\#\xi)-q^W_n(T_\#\xi)^2}$, so the GW problem is no harder than Wasserstein quantization there.
The proposed conditional-gradient algorithm with line search monotonically decreases the objective and all cluster points are stationary; with the scalar-product gauge and finite spaces, unit steps yield finite termination.

Reading between the lines

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

The two-sided bound pattern should extend to any gauge that is a Lipschitz function of a metric, so the same $n^{-1/d}$ rates likely hold for $\ell^p$ distances, bounded monotone transforms, and common similarity kernels, not just the three gauges treated here.
Because the GW objective penalizes mismatches of gauge profiles, thin structures such as limbs should collapse cross-sections before shortening along their length, which the 3D experiment displays and which could be tested quantitatively on other shapes.
Encoding a neural network as a gauged measure space suggests that pruning decisions can be made from whole-network interaction patterns rather than per-layer magnitudes, a principle that might extend to knowledge graphs, molecules, or any object with a natural pairwise gauge.

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

Read it. This is a substantial paper that does what it says: it defines GW quantization for arbitrary gauged measure spaces, proves existence and a structural characterization (block-mean gauge plus Voronoi-concentrated couplings), derives two-sided bounds against Wasserstein quantization with n^{-1/d} rates for Euclidean gauges, and provides a convergent conditional-gradient algorithm. The proofs in the appendix are detailed and use standard machinery; I didn't find a load-bearing gap.

The genuinely new part relative to Mémoli–Sidiropoulos–Singhal is the general-gauge treatment and the sharp constants. Appendix E is a nice touch: it rules out a uniform lower constant, which shows the authors know exactly what their theorems do and don't say. The Monge-map result for scalar-product spaces without absolute continuity is also a clean addition.

Where are the soft spots? The lower rate bounds require the covariance (or second-moment) matrix to have positive smallest eigenvalue, so degenerate, lower-dimensional supports are excluded. The corollaries avoid this by requiring absolute continuity, which forces full-dimensional support, so the stress-test worry is real but it's a scope limitation stated in the paper, not a flaw. The abstract's 'usual Euclidean geometries' should be read with those hypotheses in mind. The numerical rate verification uses no error bars and the pruning gauge values are hand-picked, but the experiments are clearly labeled proof-of-concept and don't carry the theory. The self-citations appear in related work and initialization; I don't see them as a problem.

The paper is for optimal-transport theorists and anyone doing clustering on non-Euclidean data. It deserves a serious referee. I'd send it to peer review.

Referee Report

0 major / 6 minor

Summary. This paper introduces Gromov–Wasserstein (GW) quantization for gauged measure spaces, in which one approximates a gm-space by a gm-space with at most n support points in GW distance. The main theoretical results are: existence and structural characterization of optimal GW quantizers (Theorem 3.4), two-sided bounds comparing the GW quantization value with the classical Wasserstein quantization value for Euclidean distance, squared Euclidean distance, and Euclidean scalar product gauges (Proposition 4.1, Theorems 4.3 and 4.5), sharp n^{-1/d} rates under absolute-continuity and moment hypotheses (Corollaries 4.4 and 4.6), existence of Monge maps for the scalar-product gauge (Theorem 4.7), a closed form in one dimension (Theorem 4.8), and a conditional-gradient/Lloyd-type algorithm with a stationarity guarantee (Theorem 5.3). The numerical section presents proof-of-concept experiments on 3D shape quantization, accelerated pairwise GW computations, rate verification, and neural-network pruning.

Significance. If correct, the paper establishes a genuine GW analogue of classical quantization theory: the n^{-1/d} rate matches the Wasserstein rate for full-dimensional absolutely continuous Euclidean measures, while the object being approximated includes the ambient gauge. The existence/characterization theorem and the Frank–Wolfe stationarity result give a principled foundation for Lloyd-type algorithms in the GW setting. The paper is honest about the scope of the rate results: the lower-bound constants depend on the measure and vanish for lower-dimensional support, and Appendix E proves that no uniform constant can exist. The proofs are detailed and follow standard compactness, L^2-mean, and Frank–Wolfe arguments; the experiments are clearly labeled as proof-of-concept and are accompanied by publicly available code.

minor comments (6)

[Abstract and Section 4 introduction] The abstract's phrase 'usual Euclidean geometries' and the opening of Section 4 suggest a universal n^{-1/d} statement, but Corollaries 4.4 and 4.6 require ξ to be absolutely continuous (and, for Corollary 4.4, X bounded), and the lower bounds in Theorems 4.3 and 4.5 degenerate when Σ_ξ or M_ξ has a zero eigenvalue. Please state these hypotheses explicitly in the abstract or in the Section 4 introduction.
[Section 6.3] The experiment uses ξ = 1/N∑δ_{x_i}, which is not absolutely continuous, so Zador's theorem and Corollaries 4.4 and 4.6 do not apply literally to the plotted values; the text's 'proxy' caveat is helpful, but the contribution bullet 'empirically confirming the quantization rates' should be weakened to 'empirically illustrating'.
[Algorithm 1 and Section 5] Theorem 5.3 assumes π_i^{(k)}(X)>0 for all i and k, while the experiments and the surrounding discussion allow t=1, which can produce empty clusters; state explicitly that the stationarity theorem applies to the line-search variant (or to positive-mass iterates) and that Corollary 5.4 covers only the scalar-product unit-step case.
[Section 4.2 after Theorem 4.5] The sharper bound using the smallest nonzero eigenvalue of M_ξ appears only in the remark after Lemma C.5; since it directly addresses the degenerate cases that the main lower bound misses, state this sharper estimate as a remark immediately after Theorem 4.5.
[Appendix C.2] The notation Var_μ(f) is used in Lemma C.2 and equation (C.2) before it is formally introduced in the sentence preceding (C.2); define it at first use in the main text or at the beginning of the appendix.
[Algorithm 1] The pseudocode computes G(π) in line 4 before π is updated in line 6 and υ in line 7; state explicitly that the Voronoi partition in line 6 is computed with the gauge G(π) of the current π, and clarify in the pseudocode itself the zero-extension convention for empty clusters.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the GW quantization rates are established by two-sided comparison with the independent Wasserstein/Zador benchmark, not by assuming the target rate.

full rationale

The paper's central rate claims (Corollaries 4.4 and 4.6) are derived by sandwiching the GW quantization value between constant multiples of the Wasserstein quantization value (Proposition 4.1, Theorems 4.3 and 4.5), with constants computed from covariance matrices, second-moment matrices, and the diameter of the space. The Wasserstein value is then controlled by Zador's theorem, an external benchmark; the paper nowhere assumes the n^{-1/d} rate for GW quantization as an input. The lower-bound proofs are self-contained: Lemma C.2 and the trace argument in Appendix C.3 reduce the GW objective to within-cell variances and compare the resulting quantity with the Wasserstein covering formulation (2.2). The existence and characterization results (Theorem 3.4) use weak compactness, the L2-mean optimality of Lemma A.2, and the bilinear relaxation of the quadratic objective; no step imports a conclusion that already encodes the theorem. The algorithm and convergence results (Proposition 5.1, Theorem 5.3, Corollary 5.4) are proved by explicit directional-derivative and Frank-Wolfe arguments. The self-citations, e.g. [8], [10], and [44], appear only in related-work remarks, a fixed-point remark of the same flavor, or as an initialization choice in the neural-network pruning experiment; none of these is load-bearing for the mathematical claims. The degenerate-covariance caveat identified by the authors, namely that the lower bound constants involve lambda_min(Sigma_xi) or lambda_min(M_xi) and vanish for lower-dimensional affine supports, is an explicitly stated scope limitation rather than a circular step; Appendix E independently constructs a family showing that no uniform lower constant can exist, which further confirms that the measure-dependent lower bounds are genuine content rather than assumed conclusions.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The theoretical core is parameter-free; the only hand-chosen numbers appear in the exploratory NN pruning experiment. The rate results rest on external Zador theory plus standard functional-analytic facts.

free parameters (2)

gauge scaling factor lambda in NN pruning encoding = 0.3
Introduced by hand in Section 6.4 to upweight short-range interactions; authors report similar results for lambda in [0.1,0.5]. Not used in the theoretical sections.
Diagonal gauge values v_i, y_l(i) in NN pruning = unspecified 'sufficiently separated' values
Chosen by hand in Section 6.4 to prevent clustering across input/output and hidden layers; only affects the proof-of-concept pruning experiment.

assumptions (6)

standard math Zador's theorem gives q_W^n(xi) asymptotically n^{-1/d} for absolutely continuous xi with finite (2+delta)-moment
Invoked in Corollaries 4.4 and 4.6 and in the introduction to Section 4.
domain assumption Gauges are symmetric square-integrable functions, so g is in L^2_sym(xi tensor xi)
Definition of gm-space in Section 3; all theory is set in this class.
domain assumption The gauge is bounded in Theorem 5.3
Required for the uniform bound on the a_k terms in the Frank-Wolfe convergence proof (Appendix D.2).
standard math Weak compactness of Pi(xi,*n) and weak continuity results (Lemma B.1)
Used throughout for existence and convergence proofs; standard results from [70] and [12].
standard math Bauer maximum principle: a convex function on a compact convex set attains its maximum at an extreme point
Central to Theorem 4.7's Monge map proof (Appendix C.4), cited from [2].
standard math Lemma A.2 (L^2 mean optimality) and Lemma A.1 (linear assignment via Voronoi)
Elementary results proved in Appendix A and used as building blocks throughout.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Gromov-Wasserstein Quantization and Clustering: Structure, Rates, and Algorithms." pith.science (2026). https://pith.science/paper/2GS7NXPI

@misc{pith2026260811016,
  author       = {Pith},
  title        = {Pith review of: Gromov-Wasserstein Quantization and Clustering: Structure, Rates, and Algorithms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2GS7NXPI}},
  note         = {Machine review of arXiv:2608.11016}
}

abstract

Clustering is a fundamental class of data analysis techniques with the most important representatives being centroid-based methods like $k$-means. Such methods are strongly connected to quantization problems, which aim to approximate general probability measures with discrete ones. For example, $k$-means corresponds to quantization with respect to the Wasserstein distance. While Wasserstein quantization clusters points within a fixed space, this paper studies Gromov-Wasserstein (GW) quantization, which additionally aims at clustering the ambient geometry of the space. We show existence of solutions to the GW quantization problem and give a characterization that justifies an analogue to the $k$-means algorithm (Lloyd's algorithm) to approximate them numerically. We further calculate the quantization rate for usual Euclidean geometries that are used in the GW context, and relate it to standard Wasserstein quantization rates. Finally, numerical experiments show that GW quantization opens up many modeling possibilities beyond normal clustering methods (e.g., for geodesic distances of 3D shapes or structured pruning of neural networks) and that the introduced algorithm leads to useful numerical solutions with approximation quality often in line with theoretically optimal rates.

Figures

Figures reproduced from arXiv: 2608.11016 by the authors.

**Figure 1.** Progressive quantization of a 3D surface for n = 1000, 500, 300, 100, 50 together with the original. Top: GW quantizers with geodesic gauge. Bottom: Wasserstein quantizers with respect to the Euclidean metric. Insets (left): a magnified view of the subject’s right forearm and hand at n = 300, with each mesh edge colored by the cluster shared by its two endpoints – GW (green frame) versus Wasserstein (orange frame). … view at source ↗

**Figure 2.** A set of 8 decimated 3D surfaces from Sumner and Popovi´c [62]. Xk ⊂ R 3 are the vertex positions, gk are the normalized pairwise Dijkstra distances of vertices within the graph induced by the mesh and ξk is the normalized area-weighted vertex measure on Xk, k = 1, . . . , 8. We consider the task of computing all pairwise distances between X1, . . . ,X8 which amounts to 8 2 = 28 GW computations. We use our quanti… view at source ↗

**Figure 3.** Total runtime (x-axis) versus mean relative error (y-axis) for approximating all 28 pairwise GW2 distances by quantization to n points. Each marker corresponds to one value of n; the dotted red line is the direct full-resolution baseline [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Verification of the two-sided bounds and the quantization rate. For each dimension d ∈ {1, 2, 3} (columns) we sample N = 5 000 points uniformly from [0, 1]d and form X = (X, g, ξ) with sampled points X, uniform measure ξ, and one of the three Euclidean gauges g (rows).…

**Figure 5.** Figure 5: Comparison of different ways to reduce the size of the neural network (three-hidden-layer feedforward neural network trained on MNIST) by reducing each hidden dimension to a width multiplier times the initial width. The baselines are pruning (gray lines, arising from …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

179 extracted references · 72 canonical work pages

[1]

Beier, Florian and von Lindheim, Johannes and Neumayer, Sebastian and Steidl, Gabriele , title =. J. Math. Imaging Vis. , volume =. 2023 , publisher =

2023
[2]

Beier, Florian and Beinert, Robert and Steidl, Gabriele , title =. Inf. Inference , volume =. 2023 , doi =

2023
[3]

IEEE Trans

Beier, Florian and Beinert, Robert and Steidl, Gabriele , title =. IEEE Trans. Image Process. , volume =. 2022 , doi =

2022
[4]

Beier, Florian and Beinert, Robert , title =. SIAM J. Imaging Sci. , volume =. 2025 , doi =

2025
[5]

Beier, Florian and Beinert, Robert , title =. Proc. Appl. Math. Mech. , volume =. 2023 , publisher =

2023
[6]

International Conference on Learning Representations (

Piening, Moritz and Beinert, Robert , title =. International Conference on Learning Representations (. 2026 , url =. 2509.22138 , archivePrefix =

arXiv 2026
[7]

International Conference on Machine Learning (

Houry, Guillaume and Feydy, Jean and Vialard, Fran. International Conference on Machine Learning (. 2026 , url =. 2602.06658 , archivePrefix =

arXiv 2026
[8]

Proceedings of the AAAI Conference on Artificial Intelligence , volume =

Piening, Moritz and Beinert, Robert , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

2026

Show all 179 references

[9]

Scale Space and Variational Methods in Computer Vision (

Beier, Florian , title =. Scale Space and Variational Methods in Computer Vision (. 2023 , publisher =

2023
[10]

Transfer Operators from Batches of Unpaired Points via Entropic Transport Kernels , journal =

Beier, Florian and Bi, Hancheng and Sarrazin, Cl. Transfer Operators from Batches of Unpaired Points via Entropic Transport Kernels , journal =. 2025 , doi =

2025
[11]

and Olkin, Ingram and Arnold, Barry C

Marshall, Albert W. and Olkin, Ingram and Arnold, Barry C. , title =. 2011 , doi =

2011
[12]

Information constrained optimal transport: From

Bai, Yikun and Wu, Xiugang and. Information constrained optimal transport: From. IEEE Trans. Inf. Theory , volume =. 2023 , publisher =

2023
[13]

Monge, Gaspard , title =. Mem. Math. Phys. Acad. Royale Sci. , pages =
[14]

, title =

Kantorovich, Leonid V. , title =. C. R. (Dokl.) Acad. Sci. URSS (N.S.) , volume =
[15]

and Slep

Wang, Wei and Ozolek, John A. and Slep. An optimal transportation approach for nuclear structure-based pathology , journal =
[16]

and Tosun, A

Ozolek, J. and Tosun, A. and Wang, W. and Chen, C. and Kolouri, S. and Basu, S. and Huang, H. and Rohde, G. , title =. Med. Image Anal. , volume =. 2014 , doi =

2014
[17]

Rubaiyat, A. H. M. and Hallam, K. M. and Nichols, J. M. and Hutchinson, M. N. and Li, S. and Rohde, G. K. , title =. IEEE Trans. Signal Process. , volume =. 2020 , doi =

2020
[18]

and Tosun, A

Kolouri, S. and Tosun, A. and Ozolek, J. and Rohde, G. , title =. Pattern Recognit. , volume =
[19]

and Kolouri, S

Park, S. and Kolouri, S. and Kundu, S. and Rohde, G. , title =. Appl. Comput. Harmon. Anal. , volume =. 2018 , doi =

2018
[20]

Emerson, T. H. and Nichols, J. M. , title =. Pattern Recognit. Lett. , volume =
[21]

and Kolouri, S

Basu, S. and Kolouri, S. and Rohde, G. , title =. Proc. Natl. Acad. Sci. USA , volume =
[22]

Tosun, A. B. and Yergiyev, O. and Kolouri, S. and Silverman, J. F. and Rohde, G. K. , title =. Cytometry
[23]

Benamou, Jean-David and Brenier, Yann , title =. Numer. Math. , volume =. 2000 , publisher =

2000
[24]

Eckermann, Marina and Schmitzer, Bernhard and van der Meer, Franziska and Franz, Jonas and Hansen, Ove and Stadelmann, Christine and Salditt, Tim , title =. Proc. Natl. Acad. Sci. USA , volume =. 2021 , publisher =

2021
[25]

, title =

McCann, Robert J. , title =. Adv. Math. , volume =. 1997 , publisher =

1997
[26]

and Taylor, Jeremy M

Kim, Dong K. and Taylor, Jeremy M. G. , title =. J. Am. Stat. Assoc. , volume =. 1995 , publisher =

1995
[27]

Jamshidian, Mortaza , title =. Comput. Stat. Data Anal. , volume =. 2004 , doi =

2004
[28]

and Kaufman, Linda , title =

Vardi, Yehuda and Shepp, Larry A. and Kaufman, Linda , title =. J. Am. Stat. Assoc. , volume =. 1985 , publisher =

1985
[29]

IEEE Trans

Byrne, Charles , title =. IEEE Trans. Image Process. , volume =. 2005 , publisher =

2005
[30]

Takai, Keiji , title =. Comput. Stat. , volume =. 2012 , doi =

2012
[31]

Pennec, Xavier , title =. J. Math. Imaging Vis. , volume =. 2006 , doi =

2006
[32]

Froyland, Gary , title =. Phys. D , volume =. 2013 , doi =

2013
[33]

Anderes, Ethan and Borgwardt, Steffen and Miller, Jacob , title =. Math. Methods Oper. Res. , volume =. 2016 , doi =

2016
[34]

Acta Math

Sturm, Karl-Theodor , title =. Acta Math. , volume =. 2006 , doi =

2006
[35]

Pattern Recognit

Nguyen, Dai Hai and Tsuda, Koji , title =. Pattern Recognit. , volume =. 2023 , publisher =

2023
[36]

Buze, Maciej , title =. SIAM J. Math. Anal. , volume =. 2025 , doi =

2025
[37]

and Ekeland, I

Carlier, G. and Ekeland, I. , title =. Econ. Theory , volume =. 2010 , publisher =

2010
[38]

and Boix-Adser

Altschuler, Jason M. and Boix-Adser. Polynomial-time algorithms for multimarginal optimal transport problems with structure , journal =. 2023 , doi =

2023
[39]

Scaling algorithms for unbalanced optimal transport problems , journal =

Chizat, L\'. Scaling algorithms for unbalanced optimal transport problems , journal =. 2018 , doi =

2018
[40]

Unbalanced optimal transport: Dynamic and

Chizat, L. Unbalanced optimal transport: Dynamic and. J. Funct. Anal. , volume =. 2018 , doi =

2018
[41]

Parallelly sliced optimal transport on spheres and on the rotation group , journal =

Quellmalz, Michael and Buecher, L. Parallelly sliced optimal transport on spheres and on the rotation group , journal =. 2024 , publisher =

2024
[42]

Delon, Julie and Desolneux, Agn. A. SIAM J. Imaging Sci. , volume =. 2020 , publisher =

2020
[43]

, title =

Zhang, Zhengxin and Goldfeld, Ziv and Mroueh, Youssef and Sriperumbudur, Bharath K. , title =. Ann. Stat. , volume =. 2024 , doi =

2024
[44]

2023 , url =

Rioux, Gabriel and Goldfeld, Ziv and Kato, Kengo , title =. 2023 , url =

2023
[45]

Inverse Probl

Quellmalz, Michael and Beinert, Robert and Steidl, Gabriele , title =. Inverse Probl. , volume =. 2023 , publisher =

2023
[46]

Nonlinearity , volume =

Junge, Oliver and Matthes, Daniel and Schmitzer, Bernhard , title =. Nonlinearity , volume =. 2024 , publisher =

2024
[47]

Optimal maps for the multidimensional

Gangbo, Wilfrid and \'. Optimal maps for the multidimensional. Commun. Pure Appl. Math. , volume =. 1998 , publisher =

1998
[48]

Agueh, Martial and Carlier, Guillaume , title =. SIAM J. Math. Anal. , volume =. 2011 , doi =

2011
[49]

Density functional theory and optimal transportation with

Cotar, Codina and Friesecke, Gero and Kl. Density functional theory and optimal transportation with. Commun. Pure Appl. Math. , volume =. 2013 , publisher =

2013
[50]

Buttazzo, Giuseppe and De Pascale, Luigi and Gori-Giorgi, Paola , title =. Phys. Rev. A , volume =. 2012 , publisher =

2012
[51]

Haasler, Isabel and Ringh, Axel and Chen, Yongxin and Karlsson, Johan , title =. SIAM J. Control Optim. , volume =. 2021 , doi =

2021
[52]

IEEE Trans

Haasler, Isabel and Singh, Rahul and Zhang, Qinsheng and Karlsson, Johan and Chen, Yongxin , title =. IEEE Trans. Inf. Theory , volume =. 2021 , publisher =

2021
[53]

A linear optimal transportation framework for quantifying and visualizing variations in sets of images , journal =

Wang, Wei and Slep. A linear optimal transportation framework for quantifying and visualizing variations in sets of images , journal =. 2013 , publisher =

2013
[54]

IEEE Trans

Zhu, Lei and Yang, Yan and Haker, Steven and Tannenbaum, Allen , title =. IEEE Trans. Image Process. , volume =. 2007 , publisher =

2007
[55]

IEEE Control Syst

Chen, Yongxin and Karlsson, Johan , title =. IEEE Control Syst. Lett. , volume =. 2018 , doi =

2018
[56]

Transfer operators from optimal transport plans for coherent set detection , journal =

Koltai, P. Transfer operators from optimal transport plans for coherent set detection , journal =. 2021 , publisher =

2021
[57]

M. Found. Comput. Math. , volume =. 2011 , doi =

2011
[58]

Proceedings of the 33rd International Conference on Machine Learning , series =

Peyr. Proceedings of the 33rd International Conference on Machine Learning , series =. 2016 , publisher =

2016
[59]

Signal Process

Elvander, Filip and Haasler, Isabel and Jakobsson, Andreas and Karlsson, Johan , title =. Signal Process. , volume =. 2020 , publisher =

2020
[60]

Optimal entropy-transport problems and a new

Liero, Matthias and Mielke, Alexander and Savar\'. Optimal entropy-transport problems and a new. Invent. Math. , volume =. 2018 , doi =

2018
[61]

Sturm, Karl-Theodor , title =. Mem. Am. Math. Soc. , volume =. 2023 , doi =

2023
[62]

2019 , note =

S. 2019 , note =. 1910.12958 , archivePrefix =

2019 arXiv
[63]

Di Marino, Simone and Gerolin, Augusto , title =. J. Sci. Comput. , volume =. 2020 , doi =

2020
[64]

From the

L. From the. J. Funct. Anal. , volume =. 2012 , publisher =

2012
[65]

Pass, Brendan , title =. J. Funct. Anal. , volume =. 2013 , publisher =

2013
[66]

On the Sequential Convergence of

Portales, L. On the Sequential Convergence of. Math. Oper. Res. , volume =. 2026 , doi =

2026
[67]

Iterative

Benamou, Jean-David and Carlier, Guillaume and Cuturi, Marco and Nenna, Luca and Peyr\'. Iterative. SIAM J. Sci. Comput. , volume =. 2015 , doi =

2015
[68]

Sliced Multi-Marginal Optimal Transport , year =

Cohen, Samuel and Terenin, Alexander and Pitcan, Yannik and Amos, Brandon and Deisenroth, Marc Peter and. Sliced Multi-Marginal Optimal Transport , year =. 2102.07115 , archivePrefix =

arXiv
[69]

Entropic regularization of

Minh, H. Entropic regularization of. J. Theor. Probab. , volume =. 2023 , doi =

2023
[70]

Vayer, Titouan and Chapel, Laetitia and Flamary, R. Fused. Algorithms , volume =. 2020 , doi =

2020
[71]

Computational optimal transport: with applications to data science , journal =

Peyr. Computational optimal transport: with applications to data science , journal =. 2019 , publisher =

2019
[72]

Chowdhury, Samir and M. The. Inf. Inference , volume =. 2019 , publisher =

2019
[73]

A fixed-point approach to barycenters in

\'. A fixed-point approach to barycenters in. J. Math. Anal. Appl. , volume =. 2016 , doi =

2016
[74]

Delon, Julie and Desolneux, Agn. J. Appl. Probab. , volume =. 2022 , doi =

2022
[75]

, title =

Lloyd, Stuart P. , title =. IEEE Trans. Inf. Theory , volume =. 1982 , doi =

1982
[76]

Asymptotics of the quantization problem on metric measure spaces , journal =

Ayd. Asymptotics of the quantization problem on metric measure spaces , journal =. 2026 , doi =

2026
[77]

and Hayase, Jonathan and Srinivasa, Siddhartha S

Ainsworth, Samuel K. and Hayase, Jonathan and Srinivasa, Siddhartha S. , title =. International Conference on Learning Representations (. 2023 , url =. 2209.04836 , archivePrefix =

2023 arXiv
[78]

von Lindheim, Johannes , title =. Comput. Optim. Appl. , volume =. 2023 , doi =

2023
[79]

1977 , doi =

Castaing, Charles and Valadier, Michel , title =. 1977 , doi =

1977
[80]

2000 , doi =

Graf, Siegfried and Luschgy, Harald , title =. 2000 , doi =

2000
[81]

Mahajan, Meena and Nimbhorkar, Prajakta and Varadarajan, Kasturi , title =. Theor. Comput. Sci. , volume =. 2012 , doi =

2012
[82]

, title =

Ng, Michael K. , title =. Pattern Recognit. , volume =. 2000 , doi =

2000
[83]

Santambrogio, Filippo , title =. Bull. Math. Sci. , volume =. 2017 , publisher =

2017
[84]

Handbook of Mathematical Models and Algorithms in Computer Vision and Imaging: Mathematical Imaging and Vision , editor =

Neumayer, Sebastian and Steidl, Gabriele , title =. Handbook of Mathematical Models and Algorithms in Computer Vision and Imaging: Mathematical Imaging and Vision , editor =. 2023 , doi =

2023
[85]

Quadrature Errors, Discrepancies, and Their Relations to Halftoning on the Torus and the Sphere , journal =

Gr. Quadrature Errors, Discrepancies, and Their Relations to Halftoning on the Torus and the Sphere , journal =. 2012 , publisher =

2012
[86]

Teuber, Tanja and Steidl, Gabriele and Gwosdek, Pascal and Schmaltz, Christian and Weickert, Joachim , title =. SIAM J. Imaging Sci. , volume =. 2011 , publisher =

2011
[87]

Park, S. R. and Cattell, L. and Nichols, J. M. and Watnik, A. and Doster, T. and Rohde, G. K. , title =. Opt. Express , volume =. 2018 , doi =

2018
[88]

, title =

Srivastava, Sanvesh and Li, Cheng and Dunson, David B. , title =. J. Mach. Learn. Res. , volume =
[89]

Ehler, M. and Gr. Curve based approximation of measures on manifolds by discrepancy minimization , journal =. 2021 , doi =

2021
[90]

Mete , title =

Dolinsky, Yan and Soner, H. Mete , title =. Probab. Theory Relat. Fields , volume =. 2014 , publisher =

2014
[91]

Computational methods for martingale optimal transport problems , journal =

Guo, Gaoyue and Ob. Computational methods for martingale optimal transport problems , journal =. 2019 , publisher =

2019
[92]

Gradient Flows in Metric Spaces and in the Space of Probability Measures , series =

Ambrosio, Luigi and Gigli, Nicola and Savar. Gradient Flows in Metric Spaces and in the Space of Probability Measures , series =. 2008 , isbn =

2008
[93]

Jordan, Richard and Kinderlehrer, David and Otto, Felix , title =. SIAM J. Math. Anal. , volume =. 1998 , doi =

1998
[94]

Otto, Felix , title =. Commun. Partial Differ. Equ. , volume =. 2001 , doi =

2001
[95]

Hertrich, Johannes and Gr. J. Math. Anal. Appl. , volume =. 2024 , publisher =

2024
[96]

Stein, Viktor and Neumayer, Sebastian and Rux, Nicolaj and Steidl, Gabriele , title =. Anal. Appl. , volume =. 2026 , doi =

2026
[97]

Beinert, Robert and Heiss, Cosmas and Steidl, Gabriele , title =. SIAM J. Imaging Sci. , volume =. 2023 , doi =

2023
[98]

On the Existence of

Dumont, Th. On the Existence of. Found. Comput. Math. , volume =. 2025 , doi =

2025
[99]

and Border, Kim C

Aliprantis, Charalambos D. and Border, Kim C. , title =. 2006 , doi =

2006
[100]

2004 , doi =

Boyd, Stephen and Vandenberghe, Lieven , title =. 2004 , doi =

2004
[101]

and Bohigas, O

Bogomolny, E. and Bohigas, O. and Schmit, C. , title =. J. Math. Phys. Anal. Geom. , volume =
[102]

2025 , note =

Aramayo Carrasco, Xavier and Nekrashevich, Maksim and Mokrov, Petr and Burnaev, Evgeny and Korotin, Alexander , title =. 2025 , note =. 2303.05978 , archivePrefix =

2025
[103]

2023 , note =

von Lindheim, Johannes and Steidl, Gabriele , title =. 2023 , note =. 2305.07071 , archivePrefix =

2023 arXiv
[104]

Aloise, Daniel and Deshpande, Amit and Hansen, Pierre and Popat, Preyas , title =. Mach. Learn. , volume =. 2009 , doi =

2009
[105]

, title =

Braides, A. , title =. 2002 , isbn =

2002
[106]

Optimal Transport: Old and New , series =

Villani, C. Optimal Transport: Old and New , series =. 2009 , doi =

2009
[107]

, title =

Bogachev, Vladimir I. , title =. 2018 , doi =

2018
[108]

2006 , isbn =

Tan, Pang-Ning and Steinbach, Michael and Kumar, Vipin , title =. 2006 , isbn =

2006
[109]

, title =

Bogachev, Vladimir I. , title =. 2007 , isbn =

2007
[110]

Pavliotis, G. A. , title =. 2014 , isbn =

2014
[111]

Altekr\". Neural. Proceedings of the 40th International Conference on Machine Learning , series =. 2023 , editor =

2023
[112]

Advances in Neural Information Processing Systems , volume =

Altschuler, Jason and Niles-Weed, Jonathan and Rigollet, Philippe , title =. Advances in Neural Information Processing Systems , volume =. 2017 , editor =

2017
[113]

International Conference on Machine Learning (

Chen, Yifan and Yao, Rentian and Yang, Yun and Chen, Jie , title =. International Conference on Machine Learning (. 2023 , publisher =

2023
[114]

and Segarra, Santiago and Uribe, C

Taveras, Carlos A. and Segarra, Santiago and Uribe, C. IEEE International Conference on Acoustics, Speech and Signal Processing (. 2026 , pages =. doi:10.1109/ICASSP55912.2026.11463976 , eprint =

2026
[115]

Distributional Reduction: Unifying Dimensionality Reduction and Clustering with

Van Assel, Hugues and Vincent-Cuaz, C. Distributional Reduction: Unifying Dimensionality Reduction and Clustering with. Transactions on Machine Learning Research , issn =. 2025 , url =

2025
[116]

Posterior Sampling Based on Gradient Flows of the

Hagemann, Paul and Hertrich, Johannes and Altekr. Posterior Sampling Based on Gradient Flows of the. International Conference on Learning Representations (. 2024 , url =

2024
[117]

Advances in Neural Information Processing Systems , volume =

Ryner, Martin and Kronqvist, Jan and Karlsson, Johan , title =. Advances in Neural Information Processing Systems , volume =. 2023 , publisher =

2023
[118]

Linear-Time

Scetbon, Meyer and Peyr. Linear-Time. Proceedings of the 39th International Conference on Machine Learning , series =. 2022 , editor =

2022
[119]

Vayer, Titouan and Flamary, R. Sliced. Advances in Neural Information Processing Systems , volume =. 2019 , publisher =

2019
[120]

Advances in Neural Information Processing Systems , volume =

Xu, Hongteng and Luo, Dixin and Carin, Lawrence , title =. Advances in Neural Information Processing Systems , volume =. 2019 , publisher =

2019
[121]

Statistical Optimal Transport via Factored Couplings , booktitle =

Forrow, Aden and H. Statistical Optimal Transport via Factored Couplings , booktitle =. 2019 , editor =

2019
[122]

Low-Rank

Scetbon, Meyer and Cuturi, Marco and Peyr. Low-Rank. Proceedings of the 38th International Conference on Machine Learning , series =. 2021 , publisher =

2021
[123]

Advances in Neural Information Processing Systems , volume =

Luise, Giulia and Salzo, Saverio and Pontil, Massimiliano and Ciliberto, Carlo , title =. Advances in Neural Information Processing Systems , volume =. 2019 , editor =

2019
[124]

Advances in Neural Information Processing Systems , volume =

Dvurechensky, Pavel and Dvinskikh, Darina and Gasnikov, Alexander and Uribe, Cesar and Nedich, Angelia , title =. Advances in Neural Information Processing Systems , volume =. 2018 , editor =

2018
[125]

Proceedings of the 36th International Conference on Machine Learning , series =

Xu, Hongteng and Luo, Dixin and Zha, Hongyuan and Carin, Lawrence , title =. Proceedings of the 36th International Conference on Machine Learning , series =. 2019 , publisher =

2019
[126]

Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , pages =

Alvarez-Melis, David and Jaakkola, Tommi , title =. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , pages =. 2018 , publisher =

2018
[127]

The Unbalanced

S. The Unbalanced. Advances in Neural Information Processing Systems , volume =. 2021 , publisher =

2021
[128]

and Nguyen, Huy and Do, Dat and Pham, Tung and Ho, Nhat , title =

Le, Khang and Le, Dung Q. and Nguyen, Huy and Do, Dat and Pham, Tung and Ho, Nhat , title =. Proceedings of the 39th International Conference on Machine Learning , series =. 2022 , editor =

2022
[129]

Advances in Neural Information Processing Systems , volume =

Cuturi, Marco , title =. Advances in Neural Information Processing Systems , volume =. 2013 , editor =

2013
[130]

2019 IEEE 10th International Conference on Software Engineering and Service Science (ICSESS) , pages =

Guan, Shujuan and Liao, Bin and Du, You and Yin, Xiangyun , title =. 2019 IEEE 10th International Conference on Software Engineering and Service Science (ICSESS) , pages =. 2019 , publisher =

2019
[131]

Static and Dynamic Texture Mixing Using Optimal Transport , booktitle =

Ferradans, Sira and Xia, Gui-Song and Peyr. Static and Dynamic Texture Mixing Using Optimal Transport , booktitle =. 2013 , editor =

2013
[132]

On the Remarkable Efficiency of

Kahl, Max and Petra, Stefania and Schn. On the Remarkable Efficiency of. Scale Space and Variational Methods in Computer Vision (. 2023 , editor =

2023
[133]

Scale Space and Variational Methods in Computer Vision (

Rabin, Julien and Peyr. Scale Space and Variational Methods in Computer Vision (. 2012 , editor =

2012
[134]

Advances in Neural Information Processing Systems , volume =

Redko, Ievgen and Vayer, Titouan and Flamary, R\'. Advances in Neural Information Processing Systems , volume =. 2020 , editor =

2020
[135]

Unbalanced

Tran, Quang Huy and Janati, Hicham and Courty, Nicolas and Flamary, R. Unbalanced. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2023 , publisher =

2023
[136]

Unsupervised Ground Metric Learning Using

Huizing, Geert-Jan and Cantini, Laura and Peyr. Unsupervised Ground Metric Learning Using. Proceedings of the 39th International Conference on Machine Learning , series =. 2022 , editor =

2022
[137]

Advances in Neural Information Processing Systems , volume =

Arbel, Michael and Korba, Anna and Salim, Adil and Gretton, Arthur , title =. Advances in Neural Information Processing Systems , volume =. 2019 , editor =

2019
[138]

Scale Space and Variational Methods in Computer Vision (

Hertrich, Johannes and Beinert, Robert and Gr. Scale Space and Variational Methods in Computer Vision (. 2023 , editor =

2023
[139]

Learning to Predict Graphs with Fused

Brogat-Motte, Luc and Flamary, R. Learning to Predict Graphs with Fused. Proceedings of the 39th International Conference on Machine Learning , series =. 2022 , editor =

2022
[140]

2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (

Chowdhury, Samir and Needham, Tom , title =. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (. 2020 , publisher =

2020
[141]

Proceedings of the 31st International Conference on Machine Learning , series =

Cuturi, Marco and Doucet, Arnaud , title =. Proceedings of the 31st International Conference on Machine Learning , series =. 2014 , editor =

2014
[142]

and Popovi

Sumner, Robert W. and Popovi. Deformation transfer for triangle meshes , journal =. 2004 , doi =

2004
[143]

International Conference on Learning Representations (

Li, Hao and Kadav, Asim and Durdanovic, Igor and Samet, Hanan and Graf, Hans Peter , title =. International Conference on Learning Representations (
[144]

AI Revolution: Research, Ethics and Society , series =

Ding, Tianqi and Xiang, Dawei and Rivas, Pablo and Dong, Liang , title =. AI Revolution: Research, Ethics and Society , series =. 2026 , publisher =

2026
[145]

Advances in Neural Information Processing Systems , volume =

Singh, Sidak Pal and Jaggi, Martin , title =. Advances in Neural Information Processing Systems , volume =. 2020 , publisher =

2020
[146]

International Conference on Machine Learning (

Morelli, Fabian and Eckstein, Stephan , title =. International Conference on Machine Learning (. 2026 , note =. 2605.22350 , archivePrefix =

2026 arXiv
[147]

Hoefler, Torsten and Alistarh, Dan and Ben-Nun, Tal and Dryden, Nikoli and Peste, Alexandra , title =. J. Mach. Learn. Res. , volume =. 2021 , url =

2021
[148]

Seber, George A. F. and Lee, Alan J. , title =. 2003 , isbn =

2003
[149]

Flamary, R. J. Mach. Learn. Res. , volume =. 2021 , url =

2021
[150]

and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, St

Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, St. Nat. Methods , volume =. 2020 , doi =

2020
[151]

, title =

Bogo, Federica and Romero, Javier and Loper, Matthew and Black, Michael J. , title =. 2014 IEEE Conference on Computer Vision and Pattern Recognition (. 2014 , publisher =

2014
[152]

and Millman, K

Harris, Charles R. and Millman, K. Jarrod and van der Walt, St. Array Programming with. Nature , volume =. 2020 , doi =

2020
[153]

, title =

Hunter, John D. , title =. Comput. Sci. Eng. , volume =. 2007 , doi =

2007
[154]

Zico , title =

Sun, Mingjie and Liu, Zhuang and Bair, Anna and Kolter, J. Zico , title =. International Conference on Learning Representations (. 2024 , url =

2024
[155]

, title =

Zador, Paul L. , title =. IEEE Trans. Inf. Theory , volume =. 1982 , doi =

1982
[156]

and Neuhoff, David L

Gray, Robert M. and Neuhoff, David L. , title =. IEEE Trans. Inf. Theory , volume =. 1998 , doi =

1998
[157]

Approximation by finitely supported measures , journal =

Kloeckner, Beno. Approximation by finitely supported measures , journal =. 2012 , doi =

2012
[158]

Introduction to vector quantization and its applications for numerics , journal =

Pag. Introduction to vector quantization and its applications for numerics , journal =. 2015 , doi =

2015
[159]

Learning Probability Measures with respect to Optimal Transport Metrics , booktitle =

Ca. Learning Probability Measures with respect to Optimal Transport Metrics , booktitle =. 2012 , publisher =

2012
[160]

and Ismail, Mohamed A

Selim, Shokri Z. and Ismail, Mohamed A. , title =. IEEE Trans. Pattern Anal. Mach. Intell. , volume =. 1984 , doi =

1984
[161]

Proceedings of the Eighteenth Annual

Arthur, David and Vassilvitskii, Sergei , title =. Proceedings of the Eighteenth Annual. 2007 , publisher =

2007
[162]

Frank, Marguerite and Wolfe, Philip , title =. Nav. Res. Logist. Q. , volume =. 1956 , doi =

1956
[163]

2016 , note =

Lacoste-Julien, Simon , title =. 2016 , note =. 1607.00345 , archivePrefix =

2016 arXiv
[164]

2007 , note =

Gromov, Misha , title =. 2007 , note =

2007
[165]

Greven, Andreas and Pfaffelhuber, Peter and Winter, Anita , title =. Probab. Theory Relat. Fields , volume =. 2009 , doi =

2009
[166]

Comparison results for

M. Comparison results for. ESAIM Control Optim. Calc. Var. , volume =. 2024 , doi =

2024
[167]

Entropic Metric Alignment for Correspondence Problems , journal =

Solomon, Justin and Peyr. Entropic Metric Alignment for Correspondence Problems , journal =. 2016 , doi =

2016
[168]

Kerdoncuff, Tanguy and Emonet, R. Sampled. Mach. Learn. , volume =. 2021 , doi =

2021
[169]

Machine Learning and Knowledge Discovery in Databases (

Chowdhury, Samir and Miller, David and Needham, Tom , title =. Machine Learning and Knowledge Discovery in Databases (. 2021 , publisher =

2021
[170]

Sketching and Clustering Metric Measure Spaces , year =

M. Sketching and Clustering Metric Measure Spaces , year =. 1801.00551 , archivePrefix =

arXiv
[171]

Proceedings of the 24th International Conference on Artificial Intelligence and Statistics , series =

Chowdhury, Samir and Needham, Tom , title =. Proceedings of the 24th International Conference on Artificial Intelligence and Statistics , series =. 2021 , publisher =

2021
[172]

Online Graph Dictionary Learning , booktitle =

Vincent-Cuaz, C. Online Graph Dictionary Learning , booktitle =. 2021 , publisher =

2021
[173]

Semi-relaxed

Vincent-Cuaz, C. Semi-relaxed. International Conference on Learning Representations (
[174]

Proceedings of the AAAI Conference on Artificial Intelligence , volume =

Xu, Hongteng , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2020 , doi =

2020
[175]

Loukas, Andreas , title =. J. Mach. Learn. Res. , volume =. 2019 , url =

2019
[176]

, title =

Bravo-Hermsdorff, Gecia and Gunderson, Lee M. , title =. Advances in Neural Information Processing Systems , volume =. 2019 , publisher =

2019
[177]

Demetci, Pinar and Santorella, Rebecca and Sandstede, Bj. J. Comput. Biol. , volume =. 2022 , doi =

2022
[178]

2022 , note =

Akash, Aditya Kumar and Li, Sixu and Garc. 2022 , note =. 2210.06671 , archivePrefix =

2022 arXiv
[179]

, title =

Han, Song and Mao, Huizi and Dally, William J. , title =. International Conference on Learning Representations (

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Beier, Florian and von Lindheim, Johannes and Neumayer, Sebastian and Steidl, Gabriele , title =. J. Math. Imaging Vis. , volume =. 2023 , publisher =

2023

[2] [2]

Beier, Florian and Beinert, Robert and Steidl, Gabriele , title =. Inf. Inference , volume =. 2023 , doi =

2023

[3] [3]

IEEE Trans

Beier, Florian and Beinert, Robert and Steidl, Gabriele , title =. IEEE Trans. Image Process. , volume =. 2022 , doi =

2022

[4] [4]

Beier, Florian and Beinert, Robert , title =. SIAM J. Imaging Sci. , volume =. 2025 , doi =

2025

[5] [5]

Beier, Florian and Beinert, Robert , title =. Proc. Appl. Math. Mech. , volume =. 2023 , publisher =

2023

[6] [6]

International Conference on Learning Representations (

Piening, Moritz and Beinert, Robert , title =. International Conference on Learning Representations (. 2026 , url =. 2509.22138 , archivePrefix =

arXiv 2026

[7] [7]

International Conference on Machine Learning (

Houry, Guillaume and Feydy, Jean and Vialard, Fran. International Conference on Machine Learning (. 2026 , url =. 2602.06658 , archivePrefix =

arXiv 2026

[8] [8]

Proceedings of the AAAI Conference on Artificial Intelligence , volume =

Piening, Moritz and Beinert, Robert , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

2026