Pith. sign in

REVIEW 3 major objections 6 minor 51 references

Under a denoising loss, linear attention can only learn an average spectral filter over the training graphs; Graph Convolutional Attention adapts to each input spectrum and closes a gap equal to spectral diversity.

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

2026-07-11 00:03 UTC pith:Q6EV2UZZ

load-bearing objection Clean denoising-loss analysis of linear vs spectrum-conditioned attention, with a usable GCA architecture; the SBM optimality equality is asymptotic and untested on the real graphs where gains are claimed. the 3 major comments →

arxiv 2607.06546 v1 pith:Q6EV2UZZ submitted 2026-07-07 cs.LG cs.AI

Graph Convolutional Attention: A Spectral Perspective on Graph Denoising and Diffusion

Shervin Khalafi , Igor Krawczuk , Sergio Rozada , Charilaos Kanatsoulis , Antonio G Marques , Alejandro Ribeiro This is my paper
classification cs.LG cs.AI
keywords graph denoisinggraph diffusionspectral attentiongraph convolutional attentionspectral diversitygraph transformersstochastic block modelseigenvector denoising
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.

The paper argues that standard linear attention is the wrong default for graph denoising. Its optimal query-key map is only the unconditional average of clean graphs expressed in the noisy eigenbasis, so it cannot track graphs whose spectra differ across the dataset. Spectral Attention conditions that map on the noisy eigenvalues and is strictly better by a quantity the authors call spectral diversity. Graph Convolutional Attention (GCA) is a practical, permutation-equivariant version: queries and keys are built from graph-filtered features, so the attention pattern is a pointwise function of the spectrum. For large stochastic block models this matches the ideal spectral map, and the softmax that follows further denoises eigenvectors by roughly projecting onto the clean principal subspace. Empirically, swapping linear projections for GCA improves denoising and diffusion, with gains that track spectral diversity, and can match full graph-transformer diffusion performance without expensive hand-crafted spectral features.

Core claim

Under Frobenius denoising loss, the best linear-attention map is the unconditional mean of B = ŨᵀAŨ, while spectral attention achieves the conditional mean given the noisy eigenvalues; the loss gap is exactly the spectral diversity of the graph distribution. GCA implements that conditional filter with graph-filtered queries and keys, asymptotically matches spectral attention on spiked SBM-type models, and softmax adds further eigenvector denoising via approximate projection onto the clean eigenspace.

What carries the argument

Spectral diversity SD(D_G) = E[‖E[B|Λ̃] − E[B]‖_F²] (with B = ŨᵀAŨ): the variance of the optimal spectral predictor. It is the exact improvement of spectral attention over linear attention, and GCA targets the same conditional map via pointwise spectral filters on the input eigenvalues.

Load-bearing premise

The proof that GCA matches ideal spectral attention assumes graphs look like large community models: a fixed low-rank community signal, within-block randomness small enough to ignore, and Wigner-type noise so the conditional map collapses to a simple pointwise outlier-shrinkage rule.

What would settle it

On synthetic SBMs with controlled block-size heterogeneity, plot GCA-versus-linear validation-loss gap against estimated spectral diversity; if the gap does not rise with diversity, or if on large spiked SBMs GCA fails to approach the theoretical spectral-attention loss, the central claim fails.

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

If this is right

  • Replacing linear query-key projections with graph-filter banks should improve denoising whenever graphs vary spectrally across the training distribution.
  • Per-dataset denoising gains should correlate with a measurable spectral-diversity surrogate.
  • In discrete graph diffusion, GCA can stay competitive while dropping costly hand-crafted spectral node features and, with random-feature encodings, explicit eigendecomposition.
  • Softmax after attention is not only a normalizer: once eigenvalues are recovered it can further denoise eigenvectors by approximate projection onto the clean principal subspace.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Spectral diversity is a practical dataset diagnostic for deciding when spectral inductive bias is worth the compute.
  • The same average-versus-conditional gap may apply to attention denoisers on other structured objects whose spectra vary (meshes, manifolds, point clouds).
  • As graph size grows, random-feature positional encodings plus GCA may become a default scalable recipe for graph diffusion because they avoid per-step eigendecomposition.

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

3 major / 6 minor

Summary. The paper analyzes attention-based graph denoising under a Frobenius objective. It shows that linear attention (outer products of linear Q/K maps on noisy eigenvectors) is optimal only as the unconditional mean E[B] with B=ŨᵀAŨ, whereas spectral attention, which may depend arbitrarily on the noisy eigenvalues, achieves the conditional mean E[B|Λ̃]; the loss gap equals a spectral-diversity term SD(D_G) from the law of total variance (Props. 1–2, Thm. 1). Graph Convolutional Attention (GCA) realizes pointwise spectral filters via graph-filtered queries/keys and is permutation-equivariant; under an asymptotic spiked-Wigner/SBM regime (Assump. 1, Thm. 2), GCA matches spectral attention via outlier shrinkage η_out. Softmax is further argued to approximate projection onto the clean principal subspace and reduce eigenvector noise under equal-block, small-noise conditions (Lem. 2, Thm. 3). Empirically, replacing linear Q/K with GCA improves denoising and often diffusion metrics, with gains correlated to a kNN estimate of spectral diversity; DiGress ablations with GCA and R-PEARL match baseline quality while reducing inference cost on larger graphs.

Significance. If the analysis holds, the paper supplies a clean spectral account of why standard linear attention is limited for graph denoising and a concrete inductive bias (spectrum-adaptive, graph-filtered Q/K) that is both theoretically motivated and practically usable in diffusion models. Strengths include: (i) Propositions 1–2 and Theorem 1 follow standard conditional-mean / total-variance arguments and cleanly identify spectral diversity as the explained-variance gap; (ii) Theorem 2 ties GCA to classical spiked-Wigner outlier shrinkage with an explicit η_out; (iii) the predicted correlation between per-dataset gains and spectral diversity is a falsifiable empirical check that the authors attempt; (iv) code is released, and the DiGress + R-PEARL integration shows a practical path to avoid expensive eigendecomposition-based features. The contribution is a useful theory-to-architecture bridge for graph diffusion, even though the strongest GCA=SA equality is scoped to asymptotic SBMs.

major comments (3)
  1. [§2.2, Theorem 2, Remark 2] §2.2, Theorem 2 / Remark 2: The claim that GCA matches the idealized spectral optimum (ℓ*_GCA=ℓ*_SA) is proved only under Assumption 1 and n→∞ spiked Wigner/SBM conditions (fixed-rank E[A], negligible Δ, Wigner or edge-flip-reduced noise, pointwise η_out in Eq. 18). The manuscript never checks whether real TUDataset graphs (PROTEINS, ENZYMES, COLLAB, DEEZER, IMDB) are even approximately in this regime, yet §4.3–4.4 and the abstract present GCA gains on those graphs as support for the spectral-attention story. Please either (a) add diagnostics (e.g., effective rank of E[A], bulk vs outlier spectrum, agreement of learned filters with η_out-like shrinkage) or (b) clearly separate in the main text what is proved for SBMs from what is a heuristic expressivity argument for general graphs, so readers do not over-read ℓ*_GCA=ℓ*_SA beyond its stated scope.
  2. [§2.2, Lemma 1; §4.1] §2.2 and Lemma 1 vs §4.1 Eqs. (26)–(27): The equivariance and SA-matching analysis restrict H_Q^{(l)}, H_K^{(l)} to constant diagonals so that attention reduces to a pointwise spectral filter in F_GCA. Experiments primarily use the richer multi-tap filter-bank form (and LayerNorm). That is a reasonable design choice, but it means the implemented GCAT is not the object for which ℓ*_GCA=ℓ*_SA is proved. State this gap explicitly near the architecture definition and, if possible, ablate diagonal/scalar filters vs full banks on at least the synthetic SBMs where the theory applies.
  3. [§4.4, Table 1; Appendix A, B.1] §4.4, Table 1 and Appendix A: Training-loss gains for GCAT are uniform, but structural MMDs worsen on COLLAB and DEEZER (and Orbit MMD on ENZYMES). DiGress ablations (Table 3) are single-run, early-stopped relative to the published 550k-step recipe, and the authors themselves call them exploratory. The main-text claim that GCA “matches standard graph-transformer performance” / improves diffusion should be tempered to match Appendix A: report uncertainty, avoid overstating parity, and discuss when lower NLL fails to translate into better samples.
minor comments (6)
  1. [§2–3] Notation: eA, eU, eΛ vs tildes in the abstract/theory mix with bA in §3; a short notation table would help.
  2. [Figure 1c; Appendix B.5] Figure 1c: report the exact correlation coefficient, noise model (Gaussian vs edge-flip), and k used for the diversity surrogate so the “strongly correlated” claim is reproducible from the figure alone.
  3. [§3, Theorem 3] Theorem 3 / Assumption 2: equal-block and E[‖E‖_F²]≪min(1,m²/log²k) are strong; a one-sentence practical takeaway (when softmax is expected to help) would help non-theory readers.
  4. [§1.1; Appendix D.3] Related work §1.1 / Appendix D: Specformer, PolyFormer, SAN, and PEARL are cited; a sharper sentence on how GCA differs (denoising objective + Q/K graph filters vs spectral features or filter parametrization for prediction) would reduce overlap concerns.
  5. [Supplementary header; B.5] Typos / polish: “Understanding Graph Denoising” appears as the supplementary title while the main title is different; arXiv id and “Preprint” headers are fine but unify titles. Occasional double spaces and “thediversity = 0” in B.5.
  6. [Abstract; Appendix B] Code link is given; ensure the released repo includes the spectral-diversity kNN estimator and DiGress ablation configs referenced in B.3–B.5 for full reproducibility.

Circularity Check

0 steps flagged

No significant circularity: optimality gap, GCA–SA match, and softmax claims are derived from the Frobenius objective, law of total variance, and RMT under stated assumptions—not forced by definition or self-citation.

full rationale

The load-bearing chain is self-contained. Propositions 1–2 are ordinary least-squares characterizations of the unconditional and conditional means of B = ŨᵀAŨ under the Frobenius denoising loss; Theorem 1 then identifies the gap with tr(Cov(E[vec(B)|Λ̃])) via the law of total variance and names that quantity spectral diversity. Naming the proved gap is not circular: SD is a distributional property (variance of conditional means), and the theorem equates the optimal SA–LA gap to it rather than defining improvement after the fact. Theorem 2 shows that under Assumption 1 (fixed-rank community signal, negligible Δ, Wigner/edge-flip noise, n→∞) the conditional mean collapses to a pointwise outlier-shrinkage map η_out, so ℓ*_GCA = ℓ*_SA in that regime; the argument cites external RMT (Benaych-Georges–Nadakuditi), not a self-citation uniqueness theorem. Softmax improvement (Theorem 3 / Lemma 2) is an analytic comparison of projection vs. basis mismatch under equal-block small-noise assumptions. Empirically, FVE is a plug-in kNN estimator of explained variance of B given Λ̃, calibrated by a permutation null, then correlated with trained GCAT–GT gaps—not the training loss itself—so the correlation is not forced by construction. Coauthor citations (PEARL, DiGress) supply engineering baselines, not premises that force the optimality claims. Weaknesses (SBM asymptotics untested on real TUDataset graphs) are assumption/scope issues, not circularity.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 3 invented entities

Core math uses standard linear algebra, conditional expectation, and spiked random-matrix theorems. Load-bearing domain assumptions are the low-rank SBM/Wigner noise model, asymptotic n, and small eigenvector leakage for softmax. Free parameters are architectural (filter taps, depth, k eigenvectors, α) and estimation choices (kNN m, noise ε). Invented conceptual entities are the Spectral Attention class, spectral diversity SD, and GCA as the equivariant practical form—not new physical particles, but new named mechanisms the claims depend on.

free parameters (5)
  • Graph-conv polynomial order / filter taps P (and L)
    Chosen by architecture (experiments use 2–3 taps); controls how well η can approximate the ideal spectral map.
  • Softmax temperature α (and idealized WV/WQ/WK tying to Λ)
    Theory needs an intermediate α*; practice uses standard attention softmax—α is not learned from first principles.
  • Top-k eigenvectors and model widths (k, d, heads, depth)
    Hand-set hyperparameters (e.g., k=16, L=3, d=128 in generative GT/GCAT runs) that affect capacity and PE quality.
  • Spectral-diversity kNN neighbors m and noise level ε
    Estimator hyperparameters (m=10, ε grid) used to produce the diversity–improvement correlation that validates the theory.
  • Diffusion training horizon and λ_E edge loss weight
    DiGress runs capped (~88k/300k steps vs 550k target); λ_E=5 and schedules are design choices affecting reported MMDs.
axioms (6)
  • standard math Law of total variance / conditional mean is L2-optimal predictor of B given Λ̃
    Used for Propositions 1–2 and Theorem 1; standard Hilbert-space fact.
  • standard math Spiked Wigner outlier eigenvalues and eigenvector overlaps (Benaych-Georges–Nadakuditi)
    Lemma 3 restates [1]; underpins Theorem 2’s η_out shrinkage.
  • domain assumption Assumption 1: Δ negligible vs E[A]; noise is Wigner (or edge-flip ≈ bias + Wigner); fixed-rank communities, n→∞
    Section 2.2; required for ℓ*_GCA = ℓ*_SA and pointwise η_out.
  • domain assumption Equal-block SBM and small noise for softmax projection improvement (Assumption 2 / Davis–Kahan control of S)
    Section 3 and Appendix C.7; without it Theorem 3’s α* existence is not established.
  • standard math Weierstrass: continuous η on compact intervals approximable by polynomials
    Lemma 1 links filter banks to FGCA.
  • ad hoc to paper Restricting H_Q^{(l)}, H_K^{(l)} to constant diagonals for equivariance in analysis (experiments use richer banks)
    Section 2.2 notes analysis uses diagonal filters while experiments use general filter banks—gap between theory class and practice.
invented entities (3)
  • Spectral Diversity SD(D_G) independent evidence
    purpose: Quantifies the loss gap between linear and spectral attention as variance of E[B|Λ̃].
    Defined in Theorem 1; central explanatory quantity for when spectrum-adaptive attention helps.
  • Spectral Attention function class F_SA no independent evidence
    purpose: Abstract attention maps eU g(eΛ) eUᵀ with arbitrary g of eigenvalues.
    Idealized class used to prove the diversity gap; not itself a deployed architecture.
  • Graph Convolutional Attention (GCA / GCAT) independent evidence
    purpose: Permutation-equivariant practical mechanism: graph-filtered queries and keys implementing pointwise spectral denoising.
    Main method contribution; validated empirically and matched to SA under SBM asymptotics.

pith-pipeline@v1.1.0-grok45 · 36176 in / 4190 out tokens · 113301 ms · 2026-07-11T00:03:10.880101+00:00 · methodology

0 comments
read the original abstract

Denoising graphs is a fundamental problem in graph learning and the core operation of graph diffusion models. Attention-based architectures like graph transformers have recently shown promise in denoising graphs. However, our principled understanding of attention-based graph denoising remains limited, making it unclear whether standard attention is the right mechanism for this task. Here we show that, under a denoising objective, linear attention is suboptimal and can only learn an average spectral denoising filter over the training distribution. This creates a fundamental limitation as graphs often vary spectrally across the distribution. To overcome this limitation, we introduce Spectral Attention, which directly utilizes the input graph spectrum and provably outperforms linear attention by a margin governed by the spectral diversity of the distribution. We then derive Graph Convolutional Attention (GCA), a practical and permutation-equivariant realization of this idea that implements spectral denoising through graph-filtered queries and keys. For stochastic block models, GCA provably matches the idealized Spectral Attention mechanism. We further show that the softmax operation, that follows the attention, provides additional denoising by approximately projecting noisy eigenvectors onto the clean eigenspace. Empirically, replacing linear attention with GCA consistently improves graph denoising and diffusion on synthetic and real datasets, with gains strongly correlated with spectral diversity. In DiGress, GCA matches standard graph-transformer performance without computing expensive structural features, and when combined with the recently proposed PEARL positional encodings, avoids explicit eigendecomposition computations resulting in faster inference without degrading quality. The code can be found here: github.com/shervinkhalafi/graph_conv_att

Figures

Figures reproduced from arXiv: 2607.06546 by Alejandro Ribeiro, Antonio G Marques, Charilaos Kanatsoulis, Igor Krawczuk, Sergio Rozada, Shervin Khalafi.

Figure 1
Figure 1. Figure 1: Quantitative and Qualitative Denoising comparison of GT and GCAT. [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Top: Reference graphs sampled from the dataset, Middle: Graphs generated using with GCAT, Bottom: Graphs generated using GT, Left: two-community SBM graphs, and Right: and ego-network graphs. DiGress Results. We evaluate on SPECTRE-SBM [27] and ENZYMES [3], two standard literature benchmarks. On SPECTRE-SBM we follow the protocol of DiGress [44]: the fixture contains 200 graphs of up to 200 nodes drawn fro… view at source ↗
Figure 3
Figure 3. Figure 3: Top: Reference graphs from the dataset, Middle: Generated with the DiGress + R-PEARL + GCAT variant. Bottom: Generated with DiGress baseline. • PROTEINS [3], from TUDataset: 1 113 protein-structure graphs in which nodes are secondary-structure elements connected when they are neighbours in the amino-acid se￾quence or in 3D space. Graph size ranges up to 620 nodes, with most graphs in the few-tens to low-hu… view at source ↗

Review history (2 revisions) →

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

51 extracted references · 51 canonical work pages · 14 internal anchors

  1. [1]

    Benaych-Georges and R

    F. Benaych-Georges and R. R. Nadakuditi. The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices, 2010

    work page 2010

  2. [2]

    D. Bo, C. Shi, L. Wang, and R. Liao. Specformer: Spectral graph neural networks meet transformers.arXiv preprint arXiv:2303.01028, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  3. [3]

    K. M. Borgwardt, C. S. Ong, S. Schönauer, S. V . N. Vishwanathan, A. J. Smola, and H.-P. Kriegel. Protein function prediction via graph kernels. InProceedings of the Thirteenth International Conference on Intelligent Systems for Molecular Biology (ISMB), pages 47–56, 2005

    work page 2005

  4. [4]

    E. Dai, W. Jin, H. Liu, and S. Wang. Towards robust graph neural networks for noisy graphs with sparse labels. InProceedings of the fifteenth ACM international conference on web search and data mining, pages 181–191, 2022

    work page 2022

  5. [5]

    Deng and B

    H. Deng and B. Liu. Gadpn: Graph adaptive denoising and perturbation networks via singular value decomposition.arXiv preprint arXiv:2601.08230, 2026

    work page arXiv 2026

  6. [6]

    V . P. Dwivedi and X. Bresson. A generalization of transformer networks to graphs, 2021

    work page 2021

  7. [7]

    Fatemi, L

    B. Fatemi, L. El Asri, and S. M. Kazemi. Slaps: Self-supervision improves structure learning for graph neural networks.Advances in Neural Information Processing Systems, 34:22667–22681, 2021

    work page 2021

  8. [8]

    Feizi, D

    S. Feizi, D. Marbach, M. Médard, and M. Kellis. Network deconvolution as a general method to distinguish direct dependencies in networks.Nature biotechnology, 31(8):726–733, 2013

    work page 2013

  9. [9]

    H. R. Frost. Denoising deterministic networks using iterative fourier transforms.arXiv preprint arXiv:2602.00790, 2026

    work page arXiv 2026

  10. [10]

    A. Garg. Graph transformers without positional encodings.arXiv preprint arXiv:2401.17791, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  11. [11]

    R. He, K. Gopinath, C. Desrosiers, and H. Lombaert. Spectral graph transformer networks for brain surface parcellation. In2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI), pages 372–376. IEEE, 2020

    work page 2020

  12. [12]

    W. Jin, Y . Ma, X. Liu, X. Tang, S. Wang, and J. Tang. Graph structure learning for robust graph neural networks. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 66–74, 2020

    work page 2020

  13. [13]

    J. Jo, D. Kim, and S. J. Hwang. Graph generation with diffusion mixture.arXiv preprint arXiv:2302.03596, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  14. [14]

    J. Jo, S. Lee, and S. J. Hwang. Score-based generative modeling of graphs via the system of stochastic differential equations. InInternational conference on machine learning, pages 10362–10383. PMLR, 2022

    work page 2022

  15. [15]

    Josephs, W

    N. Josephs, W. Li, and E. D. Kolaczyk. Network recovery from unlabeled noisy samples. In 2021 55th Asilomar Conference on Signals, Systems, and Computers, pages 1268–1273. IEEE, 2021

    work page 2021

  16. [16]

    C. I. Kanatsoulis, E. Choi, S. Jegelka, J. Leskovec, and A. Ribeiro. Learning efficient positional encodings with graph neural networks.arXiv preprint arXiv:2502.01122, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  17. [17]

    M. Karami. HiGen: Hierarchical graph generative networks. InAdvances in Neural Information Processing Systems (NeurIPS), 2023

    work page 2023

  18. [18]

    Kreuzer, D

    D. Kreuzer, D. Beaini, W. Hamilton, V . Létourneau, and P. Tossou. Rethinking graph transform- ers with spectral attention.Advances in neural information processing systems, 34:21618–21629, 2021. 12

    work page 2021

  19. [19]

    J. Li, M. Liu, H. Zhang, P. Wang, Y . Wen, L. Pan, and H. Cheng. Mask-gvae: Blind denoising graphs via partition. InProceedings of the Web Conference 2021, pages 3688–3698, 2021

    work page 2021

  20. [20]

    R. Liao, Y . Li, Y . Song, S. Wang, W. L. Hamilton, D. Duvenaud, R. Urtasun, and R. S. Zemel. Efficient graph generation with graph recurrent attention networks. InAdvances in Neural Information Processing Systems (NeurIPS), volume 32, pages 4257–4267, 2019

    work page 2019

  21. [21]

    D. Lim, J. Robinson, L. Zhao, T. Smidt, S. Sra, H. Maron, and S. Jegelka. Sign and basis invariant networks for spectral graph representation learning.arXiv preprint arXiv:2202.13013, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  22. [22]

    Y . Liu, C. Du, T. Pang, C. Li, M. Lin, and W. Chen. Graph diffusion policy optimization. In Advances in Neural Information Processing Systems (NeurIPS), 2024

    work page 2024

  23. [23]

    J. Luo, Q. Sun, H. Yuan, X. Fu, and J. Li. Robust graph learning against adversarial evasion attacks via prior-free diffusion-based structure purification. InProceedings of the ACM on Web Conference 2025, pages 2098–2110, 2025

    work page 2025

  24. [24]

    H. Lyu, Y . H. Kureh, J. Vendrow, and M. A. Porter. Learning low-rank latent mesoscale structures in networks.Nature Communications, 15(1):224, 2024

    work page 2024

  25. [25]

    J. Ma, M. He, and Z. Wei. Polyformer: Scalable node-wise filters via polynomial graph transformer. InProceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining, pages 2118–2129, 2024

    work page 2024

  26. [26]

    NetDiff: Deep Graph Denoising Diffusion for Ad Hoc Network Topology Generation

    F. Marcoccia, C. Adjih, and P. Mühlethaler. Netdiff: Deep graph denoising diffusion for ad hoc network topology generation.arXiv preprint arXiv:2410.08238, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    Martinkus, A

    K. Martinkus, A. Loukas, N. Perraudin, and R. Wattenhofer. SPECTRE: Spectral conditioning helps to overcome the expressivity limits of one-shot graph generators. InProceedings of the 39th International Conference on Machine Learning (ICML), volume 162 ofProceedings of Machine Learning Research, pages 15159–15179. PMLR, 2022

    work page 2022

  28. [28]

    Milanfar

    P. Milanfar. A tour of modern image filtering: New insights and methods, both practical and theoretical.IEEE signal processing magazine, 30(1):106–128, 2012

    work page 2012

  29. [29]

    Milanfar and M

    P. Milanfar and M. Delbracio. Denoising: a powerful building block for imaging, inverse prob- lems and machine learning.Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 383(2299), 2025

    work page 2025

  30. [30]

    A. Q. Nichol and P. Dhariwal. Improved denoising diffusion probabilistic models. InPro- ceedings of the 38th International Conference on Machine Learning (ICML), volume 139 of Proceedings of Machine Learning Research, pages 8162–8171. PMLR, 2021

    work page 2021

  31. [31]

    Technical Report: The Graph Spectral Token -- Enhancing Graph Transformers with Spectral Information

    Z. Pengmei and Z. Li. Technical report: The graph spectral token–enhancing graph transformers with spectral information.arXiv preprint arXiv:2404.05604, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  32. [32]

    Y . Qin, M. Madeira, D. Thanou, and P. Frossard. Defog: Discrete flow matching for graph generation.arXiv preprint arXiv:2410.04263, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  33. [33]

    S. Rey, V . Tenorio, S. Rozada, L. Martino, and A. G. Marques. Overparametrized deep encoder- decoder schemes for inputs and outputs defined over graphs. In2020 28th European Signal Processing Conference (EUSIPCO), pages 855–859. IEEE, 2021

    work page 2021

  34. [34]

    S. Rey, V . M. Tenorio, and A. G. Marqués. Robust graph filter identification and graph denoising from signal observations.IEEE Transactions on Signal Processing, 71:3651–3666, 2023

    work page 2023

  35. [35]

    GIST: Gauge-Invariant Spectral Transformers for Scalable Graph Neural Operators

    M. Rigotti, N. Thumiger, and T. Frick. Gist: Gauge-invariant spectral transformers for scalable graph neural operators.arXiv preprint arXiv:2603.16849, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  36. [36]

    Rozemberczki and R

    B. Rozemberczki and R. Sarkar. Characteristic functions on graphs: Birds of a feather, from statistical descriptors to parametric models. InProceedings of the 29th ACM International Conference on Information and Knowledge Management (CIKM), pages 1325–1334. ACM, 2020. 13

    work page 2020

  37. [37]

    Rudin.Principles of Mathematical Analysis

    W. Rudin.Principles of Mathematical Analysis. McGraw-Hill, 3rd edition, 1976

    work page 1976

  38. [38]

    Segarra, A

    S. Segarra, A. G. Marques, G. Mateos, and A. Ribeiro. Network topology inference from spectral templates.IEEE Transactions on Signal and Information Processing over Networks, 3(3):467–483, 2017

    work page 2017

  39. [39]

    J. Song, C. Meng, and S. Ermon. Denoising diffusion implicit models.arXiv preprint arXiv:2010.02502, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  40. [40]

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole. Score-based generative modeling through stochastic differential equations.arXiv preprint arXiv:2011.13456, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  41. [41]

    V . M. Tenorio, S. Rey, and A. G. Marques. Robust graph neural network based on graph denoising. In2023 57th Asilomar Conference on Signals, Systems, and Computers, pages 578–582. IEEE, 2023

    work page 2023

  42. [42]

    V . M. Tenorio, S. Rey, and A. G. Marques. Blind deconvolution of sparse graph signals in the presence of perturbations. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 9406–9410. IEEE, 2024

    work page 2024

  43. [43]

    Leveraging Graph Diffusion Models for Network Refinement Tasks

    P. Trivedi, R. Rossi, D. Arbour, T. Yu, F. Dernoncourt, S. Kim, N. Lipka, N. Park, N. K. Ahmed, and D. Koutra. Leveraging graph diffusion models for network refinement tasks.arXiv preprint arXiv:2311.17856, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  44. [44]

    DiGress: Discrete Denoising diffusion for graph generation

    C. Vignac, I. Krawczuk, A. Siraudin, B. Wang, V . Cevher, and P. Frossard. Digress: Discrete denoising diffusion for graph generation.arXiv preprint arXiv:2209.14734, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  45. [45]

    H. Wang, Z. Qin, Z. Sun, Q. Yang, and J. Shao. Unsupervised graph denoising via feature-driven matrix factorization.Information Sciences, 661:120156, 2024

    work page 2024

  46. [46]

    Y .-J. Wu, E. Levina, and J. Zhu. Generalized linear models with low rank effects for network data.arXiv preprint arXiv:1705.06772, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  47. [47]

    H. Xu, L. Xiang, J. Yu, A. Cao, and X. Wang. Speedup robust graph structure learning with low-rank information. InProceedings of the 30th ACM International Conference on Information & Knowledge Management, pages 2241–2250, 2021

    work page 2021

  48. [48]

    Yanardag and S

    P. Yanardag and S. V . N. Vishwanathan. Deep graph kernels. InProceedings of the 21st ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pages 1365–1374. ACM, 2015

    work page 2015

  49. [49]

    T. Yang, J. Meng, M. Zhou, Y . Yang, Y . Wang, X. Li, and Y . Tong. You can’t ignore either: Unifying structure and feature denoising for robust graph learning. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management, pages 4178–4182, 2024

    work page 2024

  50. [50]

    J. You, R. Ying, X. Ren, W. L. Hamilton, and J. Leskovec. GraphRNN: Generating realistic graphs with deep auto-regressive models. InProceedings of the 35th International Conference on Machine Learning (ICML), volume 80 ofProceedings of Machine Learning Research, pages 5694–5703. PMLR, 2018

    work page 2018

  51. [51]

    Understanding Graph Denoising

    B. Zhou, R. Li, X. Zheng, Y . G. Wang, and J. Gao. Graph denoising with framelet regularizers. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12):7606–7617, 2024. 14 Supplementary Materials for “Understanding Graph Denoising” A Limitations and Future Directions Expanding the scope of the experiments to more graph benchmark datasets, bo...

    work page arXiv 2024