Pith. sign in

REVIEW 2 major objections 6 minor 55 references

Higher-Order Group Synchronization

T0 review · 2 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper establishes that higher-order group synchronization on hypergraphs is solvable exactly when all 1-cycles of the hyperedge measurements are consistent, and it provides the first general message-passing algorithm for solving such…

desk verdict A solid higher-order extension of CEMP with a real theory–experiment gap in the robustness claims; worth refereeing after revision. read the letter →

arxiv 2505.21932 v2 pith:PAIHHI5C submitted 2025-05-28 stat.ML cs.CVcs.LGmath.COmath.OC

Adriana L. Duncan , Joe Kileel This is my paper
classification stat.MLcs.CVcs.LGmath.COmath.OC MSC 05C65
keywords higher-ordergroupsynchronizationhypergraphcycleconsistencymessagepassingrotationangularcryo-electronmicroscopyoutliercorruption
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 introduces a higher-order version of group synchronization in which the local measurements live on hyperedges: an n-wise group ratio is a coset in $G^n/\Delta$ rather than a pairwise ratio in $G$. Its central theoretical claim is Theorem 3.5: a hyperedge potential is synchronizable—that is, there exists a vertex potential compatible with all hyperedge measurements, unique up to a global group action—if and only if every 1-cycle of the hypergraph is consistent. This extends the classical cycle-consistency characterization of pairwise synchronization, and the paper argues that it matters because higher-order measurements carry more redundancy and remain usable when pairwise data is unavailable or texture-poor. On the algorithmic side, the paper proposes CHMP, a message-passing method that estimates per-hyperedge corruption levels from cycle consistency, and proves linear convergence under a good-cycle condition, plus recovery guarantees under noise and outliers. The payoff claimed is improved tolerance to corruption in rotational and angular synchronization, and comparable performance to a standard package on simulated cryo-electron microscopy data.

What carries the argument

The machinery is a hyperedge potential $\varphi$ on a hypergraph $H(V,H)$, represented after the non-canonical bijection $\tau:G^n/\Delta\to G^{n-1}$ as an $(n-1)$-tuple of group ratios, together with the cycle consistency measure $\Phi(C)=\prod_{i=1}^l \tau\circ\mathrm{Res}_{h_i\to\{v_i,v_{i+1}\}}(\varphi(h_i))$ for a 1-cycle. The algorithm runs on the bipartite cycle-hyperedge graph, updating corruption estimates $s_h(t)$ by reweighting each incident cycle with weight proportional to $\prod_{h'\in N_C\setminus\{h\}}\exp(-\beta_t s_{h'}(t))$; the fixed point relies on the identity that a good cycle exactly reveals $s_h^*$. The named condition carrying the proof is the Good Cycle Condition, which requires every hyperedge to sit in a cycle whose other hyperedges are all uncorrupted.

What would settle it

Run CHMP on a dense 3-uniform hypergraph with $m=50$ vertices under the noiseless adversarial model with corruption probability $q$ such that $\lambda\ge 1/(2n+1)\approx 1/7$, and measure $\max_h |s_h(t)-s_h^*|$ at each iteration; Theorem 6.2 predicts decay like $(2n\beta_0 r^t)^{-1}$, so a stalling error above that bound would falsify the linear convergence claim. Separately, an exhaustive search over small hypergraphs for a hyperedge potential with all 1-cycles consistent but no compatible vertex assignment would falsify Theorem 3.5.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is Theorem 3.5: for a connected hypergraph with a hyperedge potential $\varphi$ taking values in the coset spaces $G^{|h|}/\Delta$, there exists a compatible vertex potential $\rho:V\to G$, unique up to a global action of $G$, if and only if every 1-cycle $C=[v_1,\dots,v_l;h_1,\dots,h_{l-1}]_1$ satisfies $\Phi(C)=\prod_{i=1}^{l}\tau\circ \mathrm{Res}_{h_i\to\{v_i,v_{i+1}\}}(\varphi(h_i))=1$. The proof constructs the vertex potential by fixing one vertex and propagating group elements along arbitrary 1-paths, then shows path-independence from 1-cycle consistency. The same cycle-consistency signal is then repurposed as an estimator of per-hyperedge corruption: on a good cycle the consistency measure equals the corruption of the remaining hyperedge, and on a bad cycle it is controlled by the sum of the other corruptions. Those two identities carry the convergence analysis of CHMP and justify using the estimated corruption levels to derive a weighted pairwise synchronization problem.

Load-bearing premise

CHMP's guarantees require the Good Cycle Condition with $\lambda<1/(2n+1)$: every hyperedge must lie in a cycle whose other hyperedges are all uncorrupted, a strong density and low-corruption premise that fails for sparse or heavily corrupted hypergraphs.

Editorial extensions

If this is right

  • A compatible vertex potential exists for a hyperedge potential if and only if every 1-cycle is consistent, so checking cycles certifies synchronizability.
  • Under the Good Cycle Condition with $\lambda<1/(2n+1)$, CHMP recovers hyperedge corruption levels exactly in the noiseless case and approximates them with error proportional to the noise level in the noisy case.
  • Because the corruption estimates are reliable, vertex recovery can be completed by refining the hypergraph to a weighted graph and applying existing MST or spectral synchronization pipelines.
  • The numerical experiments show CHMP with the spectral refinement tolerates higher corruption rates than pairwise methods for dense hypergraphs on $SO(2)$ and $SO(3)$, and matches the ASPIRE package on simulated cryo-EM common-line data.
  • The sample complexity scales as $O(n p^{-n} q_g^{-n})$, so higher-order advantages are tied to dense hypergraphs and low corruption rates.

Reading between the lines

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

  • The benefit of CHMP is likely conditional on density: the paper's own sparse-hypergraph experiments show the pairwise baseline winning, so the advertised outlier tolerance should be expected mainly in the dense regime where good cycles are abundant.
  • The cycle-consistency characterization offers a verifiable certificate: a solver could check synchronizability of a hyperedge potential by examining a finite set of 1-cycles, enabling exact-recovery conditions for tensor-based or hypergraph-based methods.
  • In distributed synchronization, treating each locally aligned cluster as a hyperedge measurement and running CHMP on the cluster-level hypergraph could provide a principled higher-order alignment step.
  • Applying tensor decompositions directly to the refined hyperedge data, as the paper hints, may allow vertex recovery without the pairwise reduction and could preserve more of the higher-order redundancy.
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

2 major / 6 minor

Summary. The paper introduces a higher-order extension of group synchronization in which hyperedges carry n-wise ratios in G^n/Δ. The main theoretical result, Theorem 3.5, states that a hyperedge potential is synchronizable if and only if all 1-cycles are consistent, generalizing the classical cycle-consistency criterion. The authors then propose CHMP, a cycle-hyperedge message-passing algorithm that estimates hyperedge corruption levels from cycle inconsistency measures, and prove fixed-point, exact-recovery, and linear-convergence results in Theorems 6.1–6.3 under an adversarial corruption model. Numerical experiments on SO(2), SO(3), and simulated cryo-EM data compare CHMP-based pipelines with pairwise synchronization methods. The paper also states a sample-complexity result and discusses extensions to nonuniform hypergraphs.

Significance. If the theorems are correct, the main contribution is conceptual and algorithmic: it shows that higher-order synchronizability reduces to ordinary 1-cycle consistency, and it provides a first general-purpose message-passing framework for higher-order measurements. The numerical results, especially the improved outlier tolerance in dense rotational synchronization, are potentially useful, and the authors are transparent about the non-apples-to-apples comparison with CEMP and about the correlation of corruption in real data (Section 8). However, the formal convergence guarantees are proved only in a low-corruption regime that the high-corruption experiments do not satisfy; as written, the robustness claims are supported empirically rather than by Theorems 6.2 and 6.3. The definitional issues in Section 3 also need repair before the central statements can be checked. With those fixes, the paper would be a solid contribution to the synchronization literature.

major comments (2)
  1. [§6.2, §7.1, Eq. (47)] Theorems 6.2 and 6.3 assume λ < 1/(2n+1), where λ is the maximal bad-cycle fraction defined in (47), and they require the parameter constraints β0 ≤ 1/(2nλ) and r < (1−λ)/(2nλ). Under the UCMH model used in §7.1 with n=3, the expected bad-cycle fraction for a hyperedge is λ ≈ 1−(1−q)^3, since the three other hyperedges in a length-4 cycle must all be good. For q=0.2, the value used in the convergence experiment of Figure 4, this gives λ≈0.488, while the theorem requires λ<1/7≈0.143; the chosen β0=1 and r=1.2 also violate the required bounds, which for λ=0.488 would demand β0≤0.341 and r<0.175. Thus Figures 4 and 6 do not demonstrate the regime of Theorems 6.2 and 6.3, and the statement in §7 that the theory also aligns with this parameter choice is unsupported. Please report the empirical λ for each experiment, test in the proven regime, or clearly present the high-corruption results as empirical evidence outside the theorems.
  2. [§3.1–3.2, Definitions 3.3 and 3.4, Eq. (14)] The definitions of k-cycles and consistent k-cycles are not well-formed as written. Definition 3.3 specifies l−k hyperedges h1,...,h_{l−k}, yet the containment condition 'vi,vi+1,...,vi+k ∈ hi' is imposed for all i=1,...,l, and Definition 3.4's product in (14) runs over i=1,...,l using h_i, which is undefined for i>l−k unless a cyclic indexing convention is supplied. In addition, (14) says τ is the bijection G^k/Δ→G^{k−1}, but for a k-cycle the restriction Res_{h_i→{v_i,...,v_{i+k}}} takes values in G^{k+1}/Δ, so the correct map is τ:G^{k+1}/Δ→G^k; this off-by-one mismatch reappears in the definition of d_C used by Algorithm 1. These are notational issues, but they sit at the center of the paper's main synchronizability criterion and should be fixed before the statements can be checked.
minor comments (6)
  1. [Eq. (64)] The 2-product metric on SO(3)^2 is written with d_SO(3)(R1,R2) in both summands; the second summand should involve the second entries (R̃1,R̃2).
  2. [Proof of Proposition 4.7] The final chain of inequalities contains inconsistent summation indices: the sum over k=1,...,l is then written as a sum over n=1,...,l, and the term s^*_{h_i} is not defined for the running index. This display should be rewritten.
  3. [Theorem 3.5 proof, Eq. (16)] The product bounds in (16) and the following display are garbled (the notation '0Y i=l−1' does not parse), and the second 1-cycle defined for paths that share vertices uses h'_i for indices outside the stated range. The argument is recoverable, but the proof needs a careful rewrite.
  4. [Proposition 6.4] The proof applies a Chernoff bound to the indicators 1_{C∈G_h} treating them as i.i.d., but cycles in N_h share hyperedges and are not independent; also the stated constant c≥75n/16 makes the bound 1−m^{n−16c/75} vacuous when equality holds. Please either justify the independence approximation or state the result with a stricter hypothesis.
  5. [§7.2] The first paragraph contains 'tough in typical cryo-EM pipelines'; this should be 'though'.
  6. [Algorithm 4] The notation hij = arg min_{h∈H:{i,j}⊆h} sh(T) ⊆ H is confusing because arg min returns a set; please clarify that hij is the set of minimizers and that bh is chosen from this set.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 3.5 and the CHMP convergence proofs are self-contained, and the cited prior work is motivational rather than load-bearing.

full rationale

The central claim, Theorem 3.5, is proved directly from Definitions 3.2–3.4: the forward direction constructs a vertex potential by path assignment and uses 1-cycle consistency for path independence, while the converse computes Φ(C)=1 from compatibility. It does not invoke any fitted quantity or a self-citation. The CHMP convergence guarantees in Theorems 6.2 and 6.3 are proved by self-contained induction using Proposition 4.7, the Good Cycle Condition (Definition 4.4), and an explicit βt schedule; their assumptions are stated rather than imported. Algorithm 1's reweighting function f is declared as a choice ('though any nonincreasing reweighting function could be substituted'), and the β0=1, r=1.2 values are inherited from [28] as a convention, not fitted to the paper's own data. The self-references [31] and [32] appear only as application motivation (trifocal tensors and common lines) and do not carry the weight of any theorem. No 'prediction' is a renamed fitted parameter; numerical claims are explicitly separated from the theory ('we don't have a rigorous theory yet to support this behavior'). The admitted limitation that the UCMH experiments may violate the Good Cycle Condition is a scope gap, not a circular reduction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper's claims depend on a specific adversarial noise model, a particular cycle set, compactness and bi-invariance of the group metric, and a fixed vertex ordering. These are all modeling or design assumptions rather than fitted parameters. The algorithm's hyperparameters beta0, r, and T are free choices but are taken from prior work, not fitted to the target result.

free parameters (3)
  • beta0 = 1
    Inverse temperature initialization in CHMP, chosen from [28]; not fitted to data.
  • r = 1.2
    Growth rate of beta_t in CHMP, chosen from [28]; not fitted to data.
  • T = 20
    Number of iterations in experiments; chosen for the numerical tests, not required by the theory.
assumptions (4)
  • domain assumption The observation model (20) partitions hyperedges into good and bad; good hyperedges are uncorrupted up to sub-Gaussian noise and bad hyperedges are arbitrary.
    Used to define s*_h and all convergence theorems; if real data deviates from this model (e.g., correlated corruption), guarantees may not hold. The authors acknowledge this in Section 8.
  • ad hoc to paper The hypergraph is n-uniform and connected, and C is the set of (n-1)-cycles of length n+1.
    Required for the CHMP update and for the specific convergence analysis; Section 4.2. The choice of this particular cycle set is not derived from the problem but is a design decision.
  • domain assumption G is compact with a bi-invariant metric, and d_{G^{n-1}} is normalized to be at most 1.
    Needed for the cycle consistency measure and the inequalities in Propositions 4.5 and 4.7 and Theorems 6.2 and 6.3; Section 4.1.
  • domain assumption The vertices of each hyperedge are ordered so that the reduction map tau is well-defined.
    tau: G^n/Delta to G^{n-1} is non-canonical and depends on vertex ordering (Section 2.2); the paper fixes an ordering but does not analyze how the choice affects the method.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Higher-Order Group Synchronization." pith.science (2026). https://pith.science/paper/PAIHHI5C

@misc{pith2026250521932,
  author       = {Pith},
  title        = {Pith review of: Higher-Order Group Synchronization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PAIHHI5C}},
  note         = {Machine review of arXiv:2505.21932}
}
read the original abstract

Group synchronization is the problem of determining reliable global estimates from noisy local measurements on networks. The typical task for group synchronization is to assign elements of a group to the nodes of a graph in a way that respects group elements given on the edges which encode information about local pairwise relationships between the nodes. In this paper, we introduce a novel higher-order group synchronization problem which operates on a hypergraph and seeks to synchronize higher-order local measurements on the hyperedges to obtain global estimates on the nodes. Higher-order group synchronization is motivated by applications to computer vision and image processing, among other computational problems. First, we define the problem of higher-order group synchronization and discuss its mathematical foundations. Specifically, we give necessary and sufficient synchronizability conditions which establish the importance of cycle consistency in higher-order group synchronization. Then, we propose the first computational framework for general higher-order group synchronization; it acts globally and directly on higher-order measurements using a message passing algorithm. We discuss theoretical guarantees for our framework, including convergence analyses under outliers and noise. Finally, we show potential advantages of our method through numerical experiments. In particular, we show that in certain cases our higher-order method applied to rotational and angular synchronization outperforms standard pairwise synchronization methods and is more robust to outliers. We also show that our method has comparable performance on simulated cryo-electron microscopy (cryo-EM) data compared to a standard cryo-EM reconstruction package.

Figures

Figures reproduced from arXiv: 2505.21932 by the authors.

Figure 1
Figure 1. Illustration of the classical group synchronization setup (left) and the higher-order [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Illustration of a 1-cycle: [1, 2, 3, 5, 1; 12, 23, 35, 15]1 (left), a 1-cycle: [1, 2, 3, 5, 1; 123, 235, 345, 15]1 (middle), and a 2-path: (1, 2, 3, 5, 4; 123, 235, 345)1 (right). In the definitions above, hi is allowed to have any number of vertices, not just those designated by the path or cycle, however |h| is required to be at least k for every hyperedge in a k-cycle or k-path. The set of vertices v1, . . . , vl… view at source ↗
Figure 3
Figure 3. Illustration of CHG. The cycle nodes are on the left (green) and the hyperedge nodes [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Convergence results for CHMP under different levels of Gaussian noise. [PITH_FULL_IMAGE:figures/full_fig_p027_4.png]
Figure 5
Figure 5. Figure 5: Phase transitions plots for CHMP and CEMP. [PITH_FULL_IMAGE:figures/full_fig_p027_5.png]
Figure 6
Figure 6. Figure 6: Comparison of rotation recovery methods for [PITH_FULL_IMAGE:figures/full_fig_p029_6.png]
Figure 7
Figure 7. Figure 7: Comparison of CHMP and CEMP where the underlying data comes from [PITH_FULL_IMAGE:figures/full_fig_p030_7.png]
Figure 8
Figure 8. Figure 8: Example simulated cryo-EM projection images EMD-2858 (left), EMD-2811 (middle) [PITH_FULL_IMAGE:figures/full_fig_p031_8.png]
Figure 9
Figure 9. Figure 9: Example simulated projection images for EMD-2811. From left to right the simulated [PITH_FULL_IMAGE:figures/full_fig_p032_9.png]
Figure 10
Figure 10. Figure 10: Comparison of rotation recovery error for simulated cryo-EM projection images for [PITH_FULL_IMAGE:figures/full_fig_p032_10.png]
Figure 11
Figure 11. Figure 11: Comparison of angular rotation recovery methods for [PITH_FULL_IMAGE:figures/full_fig_p040_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 40 canonical work pages

  1. [1]

    Seitz, and Richard Szeliski

    Sameer Agarwal, Noah Snavely, Ian Simon, Steven M. Seitz, and Richard Szeliski. Building Rome in a day. In2009 IEEE 12th International Conference on Computer Vision, pages 72–79,

    work page

  2. [2]

    Robust and efficient camera motion synchronization via matrix decomposition

    Federica Arrigoni, Beatrice Rossi, and Andrea Fusiello. Robust and efficient camera motion synchronization via matrix decomposition. In Image Analysis and Processing–ICIAP 2015 , volume 9279, 2015. doi:10.1007/978-3-319-23231-7_40 . 33

    work page doi:10.1007/978-3-319-23231-7_40 2015

  3. [3]

    Spectral synchronization of multiple views in SE (3)

    Federica Arrigoni, Beatrice Rossi, and Andrea Fusiello. Spectral synchronization of multiple views in SE (3). SIAM Journal on Imaging Sciences , 9(4):1963–1990, 2016. doi:10.1137/ 16M1060248

    work page 1963

  4. [4]

    Bandeira, Yutong Chen, Roy R

    Afonso S. Bandeira, Yutong Chen, Roy R. Lederman, and Amit Singer. Non-unique games over compact groups and orientation estimation in cryo-EM. Inverse Problems, 36(6):064002,

    work page

  5. [5]

    Graphs and Hypergraphs

    Claude Berge. Graphs and Hypergraphs. North-Holland Publishing Company, 1973

    work page 1973

  6. [6]

    Harrington, and Michael T

    Christian Bick, Elizabeth Gross, Heather A. Harrington, and Michael T. Schaub. What are higher-order networks? SIAM Review, 65(3):686–731, 2023. doi:10.1137/21M1414024

    doi:10.1137/21m1414024 2023

  7. [7]

    Minimally connected r-uniform hypergraphs

    Mark Budden, Josh Hiller, and Andrew Penland. Minimally connected r-uniform hypergraphs. Australasian Journal of Combinatorics , 82(1):1–20, 2022

    work page 2022

  8. [8]

    Synchronization over Z2 and community detection in signed multiplex networks with constraints

    Mihai Cucuringu. Synchronization over Z2 and community detection in signed multiplex networks with constraints. Journal of Complex Networks , 3(3):469–506, 2015. doi:10.1093/ comnet/cnu050

    work page 2015

Show all 55 references
  1. [9]

    Sensor network localization by eigenvector synchronization over the euclidean group

    Mihai Cucuringu, Yaron Lipman, and Amit Singer. Sensor network localization by eigenvector synchronization over the euclidean group. ACM Transactions on Sensor Networks , 8(3), 2012. doi:10.1145/2240092.2240093

    2012

  2. [10]

    S´ ark¨ ozy

    Paul Dorbec, Sylvian Gravier, and G´ abor N. S´ ark¨ ozy. Monochromatic Hamiltoniant-tight Berge-cycles in hypergraphs. Journal of Graph Theory , pages 34–44, 2008

    2008

  3. [11]

    On the instability of relative pose estimation and RANSAC’s role

    Hongyi Fan, Joe Kileel, and Benjamin Kimia. On the instability of relative pose estimation and RANSAC’s role. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8935–8943, 2022

    2022

  4. [12]

    Condition numbers in multiview geometry, instability in relative pose estimation, and RANSAC

    Hongyi Fan, Joe Kileel, and Benjamin Kimia. Condition numbers in multiview geometry, instability in relative pose estimation, and RANSAC. arXiv preprint arXiv:2310.02719 , 2023

    2023 arXiv

  5. [13]

    Representation theoretic patterns in multi- frequency class averaging for three-dimensional cryo-electron microscopy

    Yifeng Fan, Tingran Gao, and Zhizhen Zhao. Representation theoretic patterns in multi- frequency class averaging for three-dimensional cryo-electron microscopy. Information and Inference: A Journal of the IMA , 10:723–771, 2021. doi:10.1093/imaiai/iaab012

    2021 doi

  6. [14]

    A spectral method for joint community detection and orthogonal group synchronization

    Yifeng Fan, Yuehaw Khoo, and Zhizhen Zhao. A spectral method for joint community detection and orthogonal group synchronization. SIAM Journal on Matrix Analysis and Applications , 44(2):781–821, 2023. doi:10.1137/21M1467845

    2023 doi

  7. [15]

    Three-Dimensional Electron Microscopy of Macromolecular Assemblies: Vi- sualization of Biological Molecules in Their Native State

    Joachim Frank. Three-Dimensional Electron Microscopy of Macromolecular Assemblies: Vi- sualization of Biological Molecules in Their Native State . Oxford University Press, 2006

    2006

  8. [16]

    The geometry of synchronization problems and learning group actions

    Tingran Gao, Jacek Brodzki, and Sayan Mukherjee. The geometry of synchronization problems and learning group actions. Discrete and Computational Geometry , 65, 2021. doi:10.1007/ s00454-019-00100-2 . 34

    2021

  9. [17]

    Spanning trees in random regular uniform hypergraphs

    Catherine Greenhill, Mikhail Isaev, and Gary Liang. Spanning trees in random regular uniform hypergraphs. Combinatorics, Probability and Computing , 31(1):29–53, 2022. doi:10.1017/ S0963548321000158

    2022

  10. [18]

    Jacobson, Andr´ e E

    Andr´ as Gy´ arf´ as, Michael S. Jacobson, Andr´ e E. K´ ezdy, and Jen˝ o Lehel. Odd cycles andθ- cycles in hypergraphs. Discrete Mathematics , 306:2481–2491, 2006. doi:10.1016/j.disc. 2005.12.037

    2006 doi

  11. [19]

    Representation theoretic patterns in three-dimensional cryo- electron microscopy I: The intrinsic reconsitution algorithm.Annals of Mathematics, 174:1219– 1241, 2011

    Ronny Hadani and Amit Singer. Representation theoretic patterns in three-dimensional cryo- electron microscopy I: The intrinsic reconsitution algorithm.Annals of Mathematics, 174:1219– 1241, 2011

    2011

  12. [20]

    Multiple View Geometry in Computer Vision

    Richard Hartley and Andrew Zisserman. Multiple View Geometry in Computer Vision . Cam- bridge University Press, second edition, 2004

    2004

  13. [21]

    Detector-free structure from motion

    Xingyi He, Jiaming Sun, Yifan Wang, Sida Peng, Qixing Huang, Hujun Bao, and Xiaowei Zhou. Detector-free structure from motion. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 21594–21603, 2024. doi:10.1109/CVPR52733. 2024.02040

    2024

  14. [22]

    I. N. Herstein. Abstract Algebra. Wiley, third edition, 1996

    1996

  15. [23]

    Translation synchronization via truncated least squares

    Xiangru Huang, Zhenxiao Liang, and Qixing Huang. Translation synchronization via truncated least squares. Advances in Neural Information Processing systems , 30:1459–1468, 2017

    2017

  16. [24]

    On the notion of cycles in hypergraphs

    Philippe J´ egou and Samba Ndojh Ndiaye. On the notion of cycles in hypergraphs. Discrete Mathematics, 309(6):6535–6543, 2009

    2009

  17. [25]

    GPSfM: Global projective SfM using algebraic constraints on multi-view fundamental matrices

    Yoni Kasten, Amnon Geifman, Meirav Galun, and Ronen Basri. GPSfM: Global projective SfM using algebraic constraints on multi-view fundamental matrices. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 3259–3267, 2019. doi:10.1109/CVPR.2019.00338

    2019

  18. [26]

    Minimal problems for the calibrated trifocal variety

    Joe Kileel. Minimal problems for the calibrated trifocal variety. SIAM Journal on Applied Algebra and Geometry, 1(1):575–598, 2017

    2017

  19. [27]

    Lawson, Ardan Patwardhan, Matthew L

    Catherine L. Lawson, Ardan Patwardhan, Matthew L. Baker, Corey Hryc, Eduardo Sanz Garcia, Brian P. Hudson, Ingvar Lagerstedt, Steven J. Ludtke, Grigore Pintilie, Raul Sala, John D. Westbrook, Helen M. Berman, Gerard J. Kleywegt, and Wah Chiu. EMDataBank unified data resource f...

    2015

  20. [28]

    Robust group synchronization via cycle-edge message passing

    Gilad Lerman and Yunpeng Shi. Robust group synchronization via cycle-edge message passing. Foundations of Computational Mathematics , 22:1665–1741, 2022. doi:10.1007/ s10208-021-09532-w

    2022

  21. [29]

    Fast, accurate and memory-efficient partial per- mutation synchronization

    Shaohan Li, Yunpeng Shi, and Gilad Lerman. Fast, accurate and memory-efficient partial per- mutation synchronization. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15714–15722, 2022. doi:10.1109/CVPR52688.2022.01528. 35

    2022

  22. [30]

    Efficient detection of long consistent cycles and its application to distributed synchronization

    Shaohan Li, Yunpeng Shi, and Gilad Lerman. Efficient detection of long consistent cycles and its application to distributed synchronization. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 5260–5269, 2024. doi:10.1109/CVPR52733. 2024.00503

    2024

  23. [31]

    Tensor-based synchronization and the low-rankness of the block trifocal tensor

    Daniel Miao, Gilad Lerman, and Joe Kileel. Tensor-based synchronization and the low-rankness of the block trifocal tensor. In Advances in Neural Information Processing Systems, volume 37, pages 69505–69532, 2024

    2024

  24. [32]

    Duncan, Eric J

    Tommi Muller, Adriana L. Duncan, Eric J. Verbeke, and Joe Kileel. Algebraic constraints and algorithms for common lines in cryo-EM. Biological Imaging, 4:e9, 2024

    2024

  25. [33]

    Multilinear hyperquiver representations

    Tommi Muller, Vidit Nanda, and Anna Seigal. Multilinear hyperquiver representations. Foun- dations of Computational Mathematics , 2025. doi:10.1007/s10208-025-09692-z

    2025 doi

  26. [34]

    Solving the multi-way matching problem by permutation synchronization

    Deepti Pachauri, Risi Kondor, and Vikas Singh. Solving the multi-way matching problem by permutation synchronization. In Advances in Neural Information Processing Systems , vol- ume 26, 2013

    2013

  27. [35]

    Wein, Afonso S

    Amelia Perry, Alexander S. Wein, Afonso S. Bandeira, and Ankur Moitra. Message-passing algorithms for synchronization problems over compact groups. Communications on Pure and Applied Mathematics, 71(11):2275–2322, 2018. doi:doi.org/10.1002/cpa.21750

    2018 doi

  28. [36]

    A graph partitioning ap- proach to simultaneous angular reconstitution

    Gabi Pragier, Ido Greenberg, Xiuyuan Cheng, and Yoel Shkolnisky. A graph partitioning ap- proach to simultaneous angular reconstitution. IEEE Transactions on Computational Imaging, 2(3):323–334, 2016. doi:10.1109/TCI.2016.2557076

    2016

  29. [37]

    Rosen, Luca Carlone, Afonso S

    David M. Rosen, Luca Carlone, Afonso S. Bandeira, and John J. Leonard. SE-Sync: A certifi- ably correct algorithm for synchronization over the special Euclidean group. The International Journal of Robotics Research, 38(2-3):95–125, 2019. doi:10.1177/0278364918784361

    2019 doi

  30. [38]

    Sch¨ onemann

    Peter H. Sch¨ onemann. A generalized solution of the orthogonal Procrustes problem. Psy- chometrika, 31(1):1–10, 1966. doi:10.1007/BF02289451

    1966 doi

  31. [39]

    Jacobs, Amit Singer, and Ronen Basri

    Soumyadip Sengupta, Tal Amir, Meirav Galun, Tom Goldstein, David W. Jacobs, Amit Singer, and Ronen Basri. A new rank constraint on multi-view fundamental matrices, and its applica- tion to camera location recovery. In 2017 IEEE Conference on Computer Vision and Pattern Recogni...

    2017 doi

  32. [40]

    Estimation of camera locations in highly corrupted scenarios: All about that base, no shape trouble

    Yunpeng Shi and Gilad Lerman. Estimation of camera locations in highly corrupted scenarios: All about that base, no shape trouble. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2018

    2018

  33. [41]

    Message passing least squares framework and its application to rotation synchronization

    Yunpeng Shi and Gilad Lerman. Message passing least squares framework and its application to rotation synchronization. In Proceedings of the 37th International Conference on Machine Learning, volume 119, pages 8796–8806, 2020

    2020

  34. [42]

    Scalable cluster-consistency statistics for robust multi-object matching

    Yunpeng Shi, Shaohan Li, Tyler Maunu, and Gilad Lerman. Scalable cluster-consistency statistics for robust multi-object matching. In 2021 International Conference on 3D Vision (3DV), pages 352–360, 2021. doi:10.1109/3DV53792.2021.00045. 36

    2021

  35. [43]

    Robust group synchronization via quadratic programming

    Yunpeng Shi, Cole Wyeth, and Gilad Lerman. Robust group synchronization via quadratic programming. In Proceedings of the 39th International Conference on Machine Learning , volume 162, pages 20095–20105, 2022

    2022

  36. [44]

    Viewing direction estimation in cryo-EM using synchroniza- tion

    Yoel Shkolnisky and Amit Singer. Viewing direction estimation in cryo-EM using synchroniza- tion. SIAM Journal on Imaging Sciences , 5(3):1088–1110, 2012. doi:10.1137/120863642

    2012 doi

  37. [45]

    Angular synchronization by eigenvectors and semidefinite programming

    Amit Singer. Angular synchronization by eigenvectors and semidefinite programming. Applied and Computational Harmonic Analysis , 30(1):20–36, 2011. doi:10.1016/j.acha.2010.02. 001

    2011 doi

  38. [46]

    Coifman, Fred J

    Amit Singer, Ronald R. Coifman, Fred J. Sigworth, David W. Chester, and Yoel Shkolnisky. Detecting consistent common lines in cryo-EM by voting. Journal of Structural Biology , 169(3):312–322, 2010. doi:10.1016/j.jsb.2009.11.003

    2010 doi

  39. [47]

    Three-dimensional structure determination from common lines in cryo-EM by eigenvectors and semidefinite programming

    Amit Singer and Yoel Shkolnisky. Three-dimensional structure determination from common lines in cryo-EM by eigenvectors and semidefinite programming. SIAM Journal on Imaging Sciences, 4:543–572, 2011. doi:10.1137/090767777

    2011 doi

  40. [48]

    Vector diffusion maps and the connection Laplacian

    Amit Singer and Hau-tieng Wu. Vector diffusion maps and the connection Laplacian. Commu- nications on Pure and Applied Mathematics, 65(8):1067–1144, 2012. doi:10.1002/cpa.21395

    2012 doi

  41. [49]

    Viewing angle classification of cryo-electron microscopy images using eigenvectors

    Amit Singer, Zhizhen Zhao, Yoel Shkolnisky, and Ronny Hadani. Viewing angle classification of cryo-electron microscopy images using eigenvectors. SIAM Journal on Imaging Sciences , 4(2):723–759, 2011. doi:10.1137/090778390

    2011 doi

  42. [50]

    Determination of the spatial orientation of arbi- trarily arranged identical particles of unknown structure from their projections

    Boris Vainshtein and Alexander Goncharov. Determination of the spatial orientation of arbi- trarily arranged identical particles of unknown structure from their projections. Soviet Physics Doklady, 31:278, 1986

    1986

  43. [51]

    Angular reconstitution: A posteriori assignment of projection directions for 3D reconstruction

    Marin Van Heel. Angular reconstitution: A posteriori assignment of projection directions for 3D reconstruction. Ultramicroscopy, 21(2):111–123, 1987. doi:10.1016/0304-3991(87) 90078-7

    1987 doi

  44. [52]

    Exact and stable recovery of rotations for robust syn- chronization

    Lanhui Wang and Amit Singer. Exact and stable recovery of rotations for robust syn- chronization. Information and Inference: A Journal of the IMA , 2:145–193, 2013. doi: 10.1093/imaiai/iat005

    2013 doi

  45. [53]

    Computationalcryoem/aspire- python: v0.12.0, 2023

    Garrett Wright, Joakim And´ en, Vineet Bansal, Junchao Xia, Chris Langfield, Josh Carmichael, Kris Sowattanangkul, Robbie Brook, Yunpeng Shi, Ayelet Heimowitz, Gabi Pragier, Itay Sason, Amit Moscovich, Yoel Shkolnisky, and Amit Singer. Computationalcryoem/aspire- python: v0.12...

    2023 doi

  46. [2009]

    doi:10.1109/ICCV.2009.5459148

    2009

  47. [2020]

    doi:10.1088/1361-6420/ab7d2c

    doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.