Pith. sign in

REVIEW 2 major objections 5 minor 114 references

Group-invariant affinity kernels make spectral embedding recover the quotient geometry of data with known symmetries, with better sample rates.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 01:12 UTC pith:PXZ7FCST

load-bearing objection Solid, usable theory: three G-invariant kernels converge pointwise to explicit operators on M/G with a dim(G) rate gain; free-action is the main scope limit, not a hidden flaw. the 2 major comments →

arxiv 2607.08987 v1 pith:PXZ7FCST submitted 2026-07-09 cs.LG cs.NAmath.NAmath.STstat.TH

Group Invariant Spectral Embedding

Yeari Vigder , Paulina Hoyos , David Thong , Joakim and\'en , Joe Kileel , Amit Moscovich This is my paper
classification cs.LG cs.NAmath.NAmath.STstat.TH MSC 62R0762R3058J7058J5035R02
keywords spectral embeddinggraph Laplaciangroup invariancequotient manifoldmanifold learningsample complexitySO(2)SO(3)
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Standard spectral embedding builds a graph from pairwise affinities and uses Laplacian eigenvectors as coordinates. When data is unchanged under known symmetries (rotations, for example), ordinary kernels treat symmetry-related points as unrelated and therefore cannot recover the true low-dimensional shape. This paper replaces the affinity kernel with any of three group-invariant kernels (minimize over the group, integrate over the group, or map through invariant features). For data lying on a compact Riemannian manifold M acted on freely by a compact Lie group G of isometries, the resulting graph Laplacians converge pointwise to explicit second-order operators on the quotient manifold M/G. The variance error improves because the effective dimension drops by dim(G). Experiments on SO(2)- and SO(3)-symmetric point clouds and images show that the invariant embeddings recover circles, half-circles and tori while ordinary Euclidean embeddings do not, even with large samples.

Core claim

Under a free, smooth, isometric action of a compact Lie group G on a compact Riemannian data manifold M, the random-walk graph Laplacian built from any of three G-invariant kernels (minimum, integral, or invariant-features) converges pointwise, for smooth G-invariant test functions, to an explicit second-order differential operator D on the quotient N = M/G, with variance error O_P(n^{-1/2} ε^{-1/2-(d-p)/4}) that reflects the reduced dimension d-p.

What carries the argument

Three G-invariant kernels (K_min = max_g K(x,g·y), K_int = ∫_G K(x,g·y) dη, and K_IF = K(φ(x),φ(y)) for a G-invariant feature map φ) that make the graph Laplacian descend to an operator on the quotient; the main convergence theorem identifies that operator as Δ_N minus a log-δ drift (min/integral) or the pullback of a weighted Laplacian on the image of φ (features).

Load-bearing premise

The group must act freely on the data manifold so that the space of orbits is itself a smooth manifold; if some points are fixed by nontrivial group elements the quotient becomes singular and the stated operators are no longer defined on a manifold.

What would settle it

On a data set whose true quotient geometry is known (e.g., SO(3) point clouds whose only intrinsic motion is a circle), compute the spectral embedding with each invariant kernel and with the ordinary Euclidean kernel at large n; if the invariant embeddings fail to recover a circle while the Euclidean one succeeds, or if the observed sample-complexity scaling does not improve by roughly dim(G), the central claim is false.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes spectral embedding for data with known continuous symmetries by replacing the usual affinity kernel with one of three G-invariant kernels (minimization over G, integration over G, or a G-invariant feature map). Under the hypotheses that a compact Lie group G acts smoothly, freely and isometrically on a compact Riemannian submanifold M without boundary, Theorem 3.7 proves that the associated random-walk graph Laplacians converge pointwise, for smooth G-invariant test functions, to explicit second-order operators on the quotient N = M/G (Laplace–Beltrami plus a log-δ drift for the min and integral kernels; pullback of a weighted Laplacian on the image of the feature map for the invariant-features kernel). The variance term improves from the classical O_P(n^{-1/2} ε^{-1/2-d/4}) to O_P(n^{-1/2} ε^{-1/2-(d-p)/4}), reflecting the drop in effective dimension by p = dim(G). Corollaries treat non-uniform sampling and the constant-orbit-volume case (eigenfunction correspondence). Numerical checks on SO(3) recover the predicted rate slopes; experiments on rotated point clouds, tomographic projections and spinning toys show that the invariant embeddings recover the intrinsic quotient geometry while the Euclidean kernel does not.

Significance. If the pointwise analysis holds, the work supplies a clean, kernel-level route to symmetry-aware unsupervised manifold learning that is complementary to equivariant networks and to data-augmentation arguments in supervised learning. The improved sample-complexity bound (Remark 3.11) is concrete and matches the dimension reduction one expects from quotienting by G. The three kernels cover the main practical constructions (Procrustes/alignment, group averaging, and classical invariants such as Gram matrices or bispectra), and the proofs reuse standard local-smoothing + Bernstein and Riemannian-submersion tools in a transparent way. Public code and an explicit free-action hypothesis (flagged as open for orbifolds) further strengthen the contribution. The main limitation relative to the algorithmic claim is that only pointwise consistency is proved; spectral consistency of eigenvalues/eigenvectors, which Algorithm 1 actually uses, is left open.

major comments (2)
  1. Theorem 3.7 and Corollaries 3.9–3.15 establish only pointwise convergence of L_RW f. Algorithm 1 and the experimental claims, however, rely on the low-lying eigenvectors of L_RW. Spectral consistency (eigenvalue/eigenvector convergence to those of D on N) is acknowledged as open in the Conclusion but is load-bearing for the method as stated; without it the link from the proved operator limit to the recovered embeddings remains heuristic. A short discussion of what is already known for the classical case (e.g., Calder–Trillos, Cheng–Wu) and which obstacles remain for the invariant kernels would clarify the gap.
  2. Section 5 chooses distinct hand-tuned bandwidths for different kernels (e.g., ε = 47 vs. 3000 in §5.1; ε = 0.005 vs. 3×10^{-8} in §5.2) with no common selection rule or sensitivity analysis. Because the rate statements balance bias O(ε) against a variance term that depends on ε and on effective dimension, unequal ε choices make the visual comparison of geometry recovery and the claim of better sample efficiency only partially controlled. A single cross-validated or median-heuristic rule applied uniformly, or a short ablation over ε, would make the experimental support for the rate improvement more convincing.
minor comments (5)
  1. Figure 2 caption and surrounding text: the predicted slopes under ∇f = 0 are stated as −0.75 (Euclidean) and −0.5 (invariant); the fitted Euclidean slope (−0.853) is farther from theory than the invariant ones. A brief remark on the larger sample size needed for the classical rate (already noted in the text) would help the reader interpret the discrepancy.
  2. Definition 3.6 and Lemma 4.1: the normalization Vol(G) = 1 is used throughout; a one-line reminder that Haar measure is left-invariant and unique up to scale would avoid confusion when readers compare with other conventions (e.g., bi-invariant metrics on SO(3)).
  3. Example 3.16 / Eq. (31): the factor 1/2 relating the Frobenius and bi-invariant Laplacians on SO(3) is correct but easy to miss; cross-referencing Chirikjian–Kyatkin more explicitly would help.
  4. Section 3.1 (Computation): the cost discussion for min/integral kernels is useful; a short pointer to the concrete quadrature sizes used in the experiments (m = 200 group samples in §3.5) would make reproducibility easier.
  5. Typos / notation: “And´ en” and “Shkolnisky” appear with inconsistent accents; “rotoreflection” is fine but “O(d)” vs. “SO(d)” should be checked for consistency in Example 3.3; Appendix D table is helpful and could be referenced earlier.

Circularity Check

0 steps flagged

No significant circularity: Theorem 3.7 and rates follow from classical local-smoothing/Bernstein arguments, Riemannian submersion geometry, and one published prior lemma; nothing reduces to a fit or self-definition.

full rationale

The load-bearing claim (Thm. 3.7) is a pointwise asymptotic for three explicitly constructed G-invariant kernels. The minimum-kernel case is proved from scratch in §4.1 via the Fubini quotient formula (Lem. 4.1, classical), local Gaussian smoothing on N (Lem. 4.2, App. C, standard Taylor + Ricci expansion), and Bernstein concentration; the first-order coefficient -2 arises directly from the product rule on δ. The invariant-features case reduces in §4.3 to the ordinary non-uniform graph-Laplacian theorem (Thm. 2.2) on the image manifold im ϕ, again with no free parameters. The integral-kernel case invokes Rosen et al. (2024, Thm. 11) plus the classical projection formula of Le (2001) (Lem. 3.13); the cited paper is published, parameter-free under the same free-action hypotheses, and supplies an independent proof rather than an unverified uniqueness claim. Bandwidth ε is a free asymptotic parameter, not fitted to force the claimed operator or rate. Experiments are qualitative geometry-recovery checks and do not enter the derivation. No step equates a “prediction” to its own input by construction, imports a self-uniqueness theorem, or renames a known pattern. The free-action hypothesis is stated explicitly and used throughout; it is not hidden circularity.

Axiom & Free-Parameter Ledger

2 free parameters · 6 axioms · 2 invented entities

The central theorem rests on standard Riemannian geometry of free compact group actions, classical graph-Laplacian asymptotics, and three explicit kernel constructions. No new physical entities. Free parameters are experimental bandwidths only; theory is asymptotic in ε→0, n→∞.

free parameters (2)
  • kernel bandwidth ε (experiments) = dataset-dependent (e.g. 47, 3000, 0.005, 3e-8)
    Chosen from a fixed default grid per kernel/dataset (e.g. ε=47 for min/integral on point clouds, ε=0.005 for images); not derived from theory. Affects visual embeddings but not the asymptotic claim.
  • number of group samples for min/integral kernels = m=200 in rate experiment; unspecified elsewhere
    Monte Carlo / low-discrepancy samples over G (e.g. m=200 SO(2) elements in §3.5) approximate continuous min/integral; resolution is a practical free choice.
axioms (6)
  • domain assumption G compact Lie group acts smoothly, freely, and by isometries on compact boundaryless Riemannian submanifold M ⊂ R^D
    Required for N=M/G to be a smooth Riemannian manifold and for δ to be well-defined (Thm. A.1, Def. 3.6, Thm. 3.7).
  • domain assumption Orbit-volume density δ is smooth (and positive)
    Used so that limiting operators involve ∇ log δ and so that projections of G-invariant functions remain smooth (Thm. 3.7).
  • domain assumption Data are i.i.d. from uniform (or smooth density ρ) measure on M; graph is connected
    Standard sampling model for pointwise graph-Laplacian consistency (Thm. 3.7, Cor. 3.9).
  • standard math Classical pointwise convergence of random-walk graph Laplacians to Laplace–Beltrami / Fokker–Planck operators (Singer 2006; Coifman–Lafon 2006)
    Used as black-box for Euclidean and IF kernels and as template for min-kernel asymptotics.
  • standard math Rosen et al. (2024) Thm. 11: normalized G-GL converges to Δ_M with rate involving (d−p)
    Integral-kernel case reduces to this result plus projection of Δ_M (Lemma 3.13).
  • domain assumption For IF kernel, induced map ϕ̄ : N → im ϕ is a diffeomorphism
    Stated in Thm. 3.7 item 3; needed so the limiting operator lives on a manifold diffeomorphic to N.
invented entities (2)
  • minimum kernel K_min independent evidence
    purpose: G-invariant affinity via global alignment min_α ∥x−α·y∥²
    Standard construction (Procrustes / alignment); not a new physical entity. Analysis of its graph Laplacian on the quotient is new.
  • orbit-volume density δ and induced δ̄ on N independent evidence
    purpose: Measures relative orbit volumes; appears as drift term in limiting operators
    Standard in Riemannian submersion / Helgason; not invented here, but central to the explicit form of D.

pith-pipeline@v1.1.0-grok45 · 33902 in / 3637 out tokens · 36547 ms · 2026-07-13T01:12:33.629626+00:00 · methodology

0 comments
read the original abstract

Spectral embedding methods are widely used for dimensionality reduction and clustering of high-dimensional datasets with intrinsic low-dimensional structures. Although many datasets of practical interest exhibit invariance under symmetries such as rotations, standard spectral embedding methods do not account for this, treating symmetry-related data points as unrelated. Our approach to this problem is to incorporate the symmetries directly into the affinity kernels used for spectral embedding. We analyze the case of a Riemannian data manifold $M$ with symmetries given by a compact Lie group~$G$ and prove that, under suitable conditions, graph Laplacians constructed from three types of invariant kernels converge pointwise to explicit second-order differential operators on the quotient space $M/G$. Our analysis implies improved convergence rates, as the effective dimension drops according to the dimension of the group. We validate our approach on datasets with $\mathrm{SO}(2)$ or $\mathrm{SO}(3)$ symmetry, and show that $G$-invariant spectral embedding recovers the intrinsic geometry of the data, in contrast to standard spectral embedding, which fails to do so even in the limit of infinite data.

Figures

Figures reproduced from arXiv: 2607.08987 by Amit Moscovich, David Thong, Joakim And\'en, Joe Kileel, Paulina Hoyos, Yeari Vigder.

Figure 1
Figure 1. Figure 1: An illustration of group orbits and the quotient manifold. The flat torus [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Error between 4 ε (LRW f)(I3) and 1 versus ε for the G-invariant function f(R) = cos β(R) using the Euclidean, minimum, and integral kernels. Shaded regions show the 95% confidence intervals across 1000 trials. 4 Proof of Theorem 3.7 We split the proof of Theorem 3.7 into three parts, corresponding to the three G-invariant kernels considered in the statement of the theorem: the minimum kernel (Section 4.1)… view at source ↗
Figure 3
Figure 3. Figure 3: The Glucagon molecule as a point cloud (the displayed sticks are for visualization purposes): [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Spectral embedding of clean and noisy 3D point clouds of the Glucagon molecule with [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Spectral embedding of clean and noisy tomographic images based on the Euclidean vs. [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Spinning toys experiment. (a) A 2D image of two independently rotating 3D toy figurines. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

114 extracted references · 12 canonical work pages

  1. [1]

    Journal of Computer and System Sciences , volume =

    Towards a theoretical foundation for. Journal of Computer and System Sciences , volume =. 2008 , issn =. doi:10.1016/j.jcss.2007.08.006 , author =

    doi:10.1016/j.jcss.2007.08.006 2008

  2. [2]

    International Journal of Mathematics and Mathematical Sciences , volume =

    Le, Huiling , title =. International Journal of Mathematics and Mathematical Sciences , volume =. doi:10.1155/S0161171201003751 , abstract =

    work page doi:10.1155/s0161171201003751

  3. [3]

    and Prli

    Rose, Peter W. and Prli. The. Nucleic Acids Research , year =. doi:10.1093/nar/gkw1000 , pmid =

    work page doi:10.1093/nar/gkw1000

  4. [4]

    A variational approach to the consistency of spectral clustering , journal =

    Nicol. A variational approach to the consistency of spectral clustering , journal =. 2016 , volume =

    2016

  5. [5]

    2026 , organization=

    Fraiman, Michael and Hoyos, Paulina and Bendory, Tamir and Kileel, Joe and Mickelin, Oscar and Sharon, Nir and Singer, Amit , booktitle=. 2026 , organization=

    2026

  6. [6]

    IEEE Transactions on Computational Imaging , year =

    Zhao, Zhizhen and Shkolnisky, Yoel and Singer, Amit , title =. IEEE Transactions on Computational Imaging , year =. doi:10.1109/TCI.2016.2514700 , pmid =

    work page doi:10.1109/tci.2016.2514700 2016

  7. [7]

    Acta Crystallographica Section A , year =

    Wolfgang Kabsch , title =. Acta Crystallographica Section A , year =

  8. [8]

    Maslen , title =

    David K. Maslen , title =. Modern Signal Processing , editor =. 2004 , doi =

    2004

  9. [9]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

    Shinji Umeyama , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

  10. [10]

    Trefethen and J

    Lloyd N. Trefethen and J. A. C. Weideman , title =. SIAM Review , year =

  11. [11]

    Peter H. Sch. A generalized solution of the orthogonal. Psychometrika , year =

  12. [12]

    Error Estimates for Spectral Convergence of the Graph Laplacian on Random Geometric Graphs Toward the

    Nicol. Error Estimates for Spectral Convergence of the Graph Laplacian on Random Geometric Graphs Toward the. Foundations of Computational Mathematics , year =

  13. [13]

    Lowe , title =

    David G. Lowe , title =. International Journal of Computer Vision , year =

  14. [14]

    2003 , publisher=

    Introduction to Smooth Manifolds , author=. 2003 , publisher=

    2003

  15. [15]

    2021 , month =

    Nicholas Kayban , title =. 2021 , month =

    2021

  16. [16]

    Marc Alexa , booktitle=. Super-. 2022 , pages=

    2022

  17. [17]

    Eigen-convergence of

    Xiuyuan Cheng and Nan Wu , journal=. Eigen-convergence of. 2022 , volume=

    2022

  18. [18]

    1960 , publisher=

    The classification of G-spaces , authors=. 1960 , publisher=

    1960

  19. [20]

    Applied and Computational Harmonic Analysis , volume =

    The. Applied and Computational Harmonic Analysis , volume =. 2024 , issn =. doi:10.1016/j.acha.2024.101637 , author =

    work page doi:10.1016/j.acha.2024.101637 2024

  20. [21]

    doi:10.1090/surv/083 , year=

    Groups and Geometric Analysis: Integral Geometry, Invariant Differential Operators, and Spherical Functions , author=. doi:10.1090/surv/083 , year=

    doi:10.1090/surv/083

  21. [22]

    2023 , doi =

    An introduction to optimization on smooth manifolds , author =. 2023 , doi =

    2023

  22. [23]

    , title =

    Abraham, Ralph and Marsden, Jerrold E. , title =

  23. [24]

    On Spectral Clustering: Analysis and an algorithm , url =

    Ng, Andrew and Jordan, Michael and Weiss, Yair , booktitle =. On Spectral Clustering: Analysis and an algorithm , url =

  24. [25]

    Inverse Problems , abstract =

    Aaditya Rangan and Marina Spivak and Joakim Andén and Alex Barnett , title =. Inverse Problems , abstract =. 2020 , month =. doi:10.1088/1361-6420/ab4e66 , url =

    doi:10.1088/1361-6420/ab4e66 2020

  25. [26]

    2013 , publisher=

    Zhao, Zhizhen and Singer, Amit , journal=. 2013 , publisher=

    2013

  26. [27]

    Rotationally invariant image representation for viewing direction classification in cryo-

    Zhao, Zhizhen and Singer, Amit , journal=. Rotationally invariant image representation for viewing direction classification in cryo-. 2014 , publisher=

    2014

  27. [28]

    Neural Information Processing Systems (NeurIPS) , year=

    Fixed-rank approximation of a positive-semidefinite matrix from streaming data , author=. Neural Information Processing Systems (NeurIPS) , year=

  28. [29]

    SIAM Journal on Scientific computing , volume=

    An algorithm for the principal component analysis of large data sets , author=. SIAM Journal on Scientific computing , volume=. 2011 , publisher=

    2011

  29. [30]

    Applied and Computational Harmonic Analysis , volume =

    From graph to manifold. Applied and Computational Harmonic Analysis , volume =. 2006 , issn =. doi:10.1016/j.acha.2006.03.004 , author =

    doi:10.1016/j.acha.2006.03.004 2006

  30. [31]

    SIAM Journal on Imaging Sciences , volume =

    Landa, Boris and Shkolnisky, Yoel , title =. SIAM Journal on Imaging Sciences , volume =. 2018 , doi =

    2018

  31. [32]

    , title =

    Chen, Shuxiao and Dobriban, Edgar and Lee, Jane H. , title =. J. Mach. Learn. Res. , month =. 2020 , issue_date =

    2020

  32. [33]

    Laplacian Eigenmaps for Dimensionality Reduction and Data Representation , year=

    Belkin, Mikhail and Niyogi, Partha , journal=. Laplacian Eigenmaps for Dimensionality Reduction and Data Representation , year=

  33. [34]

    Tenenbaum and Vin de Silva and John C

    Joshua B. Tenenbaum and Vin de Silva and John C. Langford , title =. Science , volume =. 2000 , doi =

    2000

  34. [35]

    and Wintgen, P

    Sulanke, R. and Wintgen, P. , lccn=. Differentialgeometrie und Faserb. 1972 , publisher=

    1972

  35. [36]

    2023 , publisher=

    An Introduction to Optimization on Smooth Manifolds , author=. 2023 , publisher=

    2023

  36. [37]

    Absil and J\'

    P.-A. Absil and J\'. Projection-like Retractions on Matrix Manifolds , publisher =. 2012 , journal =

    2012

  37. [38]

    1987 , series =

    Transformation Groups , author =. 1987 , series =

    1987

  38. [39]

    Convergence of

    Belkin, Mikhail and Niyogi, Partha , booktitle =. Convergence of

  39. [40]

    2006 , issn =

    Diffusion maps , journal =. 2006 , issn =. doi:10.1016/j.acha.2006.04.006 , author =

    doi:10.1016/j.acha.2006.04.006 2006

  40. [41]

    Improved spectral convergence rates for graph

    Calder, Jeff and Trillos, Nicolas Garcia , journal=. Improved spectral convergence rates for graph. 2022 , publisher=

    2022

  41. [42]

    Methods in Enzymology , volume=

    Cryo-electron tomography and subtomogram averaging , author=. Methods in Enzymology , volume=. 2016 , publisher=

    2016

  42. [43]

    Manifold Learning with Arbitrary Norms

    Joe Kileel and Amit Moscovich and Nathan Zelesko and Amit Singer. Manifold Learning with Arbitrary Norms. Journal of Fourier Analysis and Applications. 2021. doi:10.1007/s00041-021-09879-2

    work page doi:10.1007/s00041-021-09879-2 2021

  43. [44]

    Proceedings of the Twelth International Conference on Artificial Intelligence and Statistics , pages =

    Multi-Manifold Semi-Supervised Learning , author =. Proceedings of the Twelth International Conference on Artificial Intelligence and Statistics , pages =. 2009 , editor =

    2009

  44. [45]

    2017 , editor =

    Moscovich, Amit and Jaffe, Ariel and Boaz, Nadler , booktitle =. 2017 , editor =

    2017

  45. [46]

    Journal of the American Statistical Association , volume =

    Mingyen Cheng and Hautieng Wu , title =. Journal of the American Statistical Association , volume =. 2013 , publisher =

    2013

  46. [47]

    1995 , publisher=

    Group Theory and Physics , author=. 1995 , publisher=

    1995

  47. [48]

    2016 , publisher=

    Harmonic Analysis for Engineers and Applied Scientists: Updated and Expanded Edition , author=. 2016 , publisher=

    2016

  48. [49]

    Approximation of functions over manifolds: A Moving Least-Squares approach , journal =

    Barak Sober and Yariv Aizenbud and David Levin , keywords =. Approximation of functions over manifolds: A Moving Least-Squares approach , journal =. 2021 , issn =. doi:10.1016/j.cam.2020.113140 , url =

    work page doi:10.1016/j.cam.2020.113140 2021

  49. [50]

    The Journal of Machine Learning Research , volume=

    A group-theoretic framework for data augmentation , author=. The Journal of Machine Learning Research , volume=. 2020 , publisher=

    2020

  50. [51]

    Annual Review of Biomedical Data Science , volume=

    Computational methods for single-particle electron cryomicroscopy , author=. Annual Review of Biomedical Data Science , volume=. 2020 , publisher=

    2020

  51. [52]

    Acta Crystallographica Section D: Structural Biology , volume=

    High-resolution structure determination using high-throughput electron cryo-tomography , author=. Acta Crystallographica Section D: Structural Biology , volume=. 2022 , publisher=

    2022

  52. [53]

    2011 , publisher=

    Stochastic Models, Information Theory, and Lie Groups, Volume 2: Analytic Methods and Modern Applications , author=. 2011 , publisher=

    2011

  53. [54]

    2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI) , pages=

    Earthmover-based manifold learning for analyzing molecular conformation spaces , author=. 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI) , pages=. 2020 , organization=

    2020

  54. [55]

    The Laplacian on a Riemannian Manifold: An Introduction to Analysis on Manifolds , publisher=

    Rosenberg, Steven , year=. The Laplacian on a Riemannian Manifold: An Introduction to Analysis on Manifolds , publisher=

  55. [56]

    2008 , publisher=

    Group Theoretical Methods in Machine Learning , author=. 2008 , publisher=

    2008

  56. [57]

    1946 , edition =

    Hermann Weyl , title =. 1946 , edition =

    1946

  57. [58]

    2020 , edition =

    Jean Gallier and Jocelyn Quaintance , title =. 2020 , edition =. doi:10.1007/978-3-030-46040-2 , note =

    doi:10.1007/978-3-030-46040-2 2020

  58. [59]

    2004 , edition =

    Multiple View Geometry in Computer Vision , author =. 2004 , edition =

    2004

  59. [60]

    Journal of Machine Learning Research , volume =

    Visualizing Data using t-SNE , author =. Journal of Machine Learning Research , volume =. 2008 , url =

    2008

  60. [61]

    Reinhart , journal =

    Bruce L. Reinhart , journal =. Foliated Manifolds with Bundle-Like Metrics , urldate =. 1959 , doi =

    1959

  61. [62]

    and Weizsäcker, H.v

    Smolyanov, O.G. and Weizsäcker, H.v. and Wittich, O. , title =. Potential Analysis , volume =. 2007 , doi =

    2007

  62. [63]

    , title =

    Sherbert, Donald R. , title =. Pacific Journal of Mathematics , volume =. 1963 , doi =

    1963

  63. [64]

    Sigurdur Helgason , title =. J. Differential Geometry , volume =. 1972 , pages =

    1972

  64. [65]

    Lee , title =

    John M. Lee , title =. 2018 , edition =. doi:10.1007/978-3-319-91755-9 , pages =

    doi:10.1007/978-3-319-91755-9 2018

  65. [66]

    2017 , edition =

    Jost, Jürgen , title =. 2017 , edition =

    2017

  66. [67]

    2009 , series =

    Grigor'yan, Alexander , title =. 2009 , series =

    2009

  67. [68]

    2014 , series =

    Bakry, Dominique and Gentil, Ivan and Ledoux, Michel , title =. 2014 , series =

    2014

  68. [69]

    1996 , edition =

    Risken, Hannes , title =. 1996 , edition =

    1996

  69. [70]

    1996 , isbn =

    Sakai, Takashi , title =. 1996 , isbn =

    1996

  70. [71]

    and Tickle, Ian J

    Sasaki, Kyoyu and Dockerill, Susan and Adamiak, Dorota A. and Tickle, Ian J. and Blundell, Tom , year =. X-ray analysis of glucagon and its relationship to receptor binding , volume =. Nature , publisher =. doi:10.1038/257751a0 , number =

    work page doi:10.1038/257751a0

  71. [72]

    and Sasaki, K

    Blundell, T.L. and Sasaki, K. and Dockerill, S. and Tickle, I.J. , year =. X-RAY ANALYSIS OF GLUCAGON AND ITS RELATIONSHIP TO RECEPTOR BINDING , url =. doi:10.2210/pdb1gcn/pdb , publisher =

    work page doi:10.2210/pdb1gcn/pdb

  72. [73]

    2025 , copyright =

    Garrett Wright and Joakim Andén and Vineet Bansal and Junchao Xia and Chris Langfield and Josh Carmichael and Kris Sowattanangkul and Robbie Brook and Yunpeng Shi and Ayelet Heimowitz and Gabi Pragier and Itay Sason and Amit Moscovich and Yoel Shkolnisky and Amit Singer , title =. 2025 , copyright =. doi:10.5281/ZENODO.5657281 , url =

    work page doi:10.5281/zenodo.5657281 2025

  73. [74]

    Statistics and Computing , volume =

    von Luxburg, Ulrike , title =. Statistics and Computing , volume =. 2007 , doi =

    2007

  74. [75]

    The Annals of Statistics , year =

    von Luxburg, Ulrike and Belkin, Mikhail and Bousquet, Olivier , title =. The Annals of Statistics , year =

  75. [76]

    International Conference on Machine Learning (ICML) , year =

    Group Equivariant Convolutional Networks , author =. International Conference on Machine Learning (ICML) , year =

  76. [77]

    Advances in Neural Information Processing Systems (NeurIPS) , volume =

    Deep Sets , author =. Advances in Neural Information Processing Systems (NeurIPS) , volume =. 2017 , url =

    2017

  77. [78]

    and Welling, Max , booktitle =

    Cohen, Taco S. and Welling, Max , booktitle =. Steerable. 2017 , url =

    2017

  78. [79]

    International Conference on Learning Representations (ICLR) , year =

    Invariant and Equivariant Graph Networks , author =. International Conference on Learning Representations (ICLR) , year =

  79. [80]

    arXiv preprint arXiv:2104.13478 , year =

    Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges , author =. arXiv preprint arXiv:2104.13478 , year =

    Pith/arXiv arXiv

  80. [81]

    Communications on Pure and Applied Mathematics , volume =

    Singer, Amit and Wu, Hau-Tieng , title =. Communications on Pure and Applied Mathematics , volume =. 2012 , doi =

    2012

Showing first 80 references.