Sheaf-Laplacian Obstruction and Projection Hardness for Cross-Modal Compatibility on a Modality-Independent Site

Tibor Sloboda

REVIEW 4 major objections 5 minor 18 references

Sheaf-Laplacian Obstruction and Projection Hardness for Cross-Modal Compatibility on a Modality-Independent Site

T0 review · 4 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read Cross-modal compatibility is made measurable by two invariants—projection hardness and sheaf-Laplacian obstruction—and is provably non-transitive.

desk verdict A useful conceptual framework for cross-modal alignment, but the central ReLU non-transitivity construction does not hold under the paper's own whitening and Lipschitz constraints. read the letter →

arxiv 2604.07632 v2 pith:J6OUKN24 submitted 2026-04-08 cs.LG cs.AI

Tibor Sloboda This is my paper · ORCID

classification cs.LGcs.AI

keywords cross-modalalignmentprojectionhardnesssheaf-Laplacianobstructionmodality-independentsitenon-transitivecompatibilitybridgemodalitygraphspectralgapsheaf-regularizedregression

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

This paper tries to make cross-modal alignment a well-defined, comparable quantity by evaluating every modality on a single fixed neighborhood graph over samples, then reading off two directed invariants: projection hardness H, the smallest complexity of one global projection that reaches a target error, and sheaf-Laplacian obstruction C, the smallest spatial variation in locally fit projection parameters that reaches the same error. The central payoff is formal: C is computed by an exactly implemented sheaf-regularized regression objective, relates to the spectral gap of the graph, and captures failures that global expressivity alone cannot explain. In a one-dimensional ReLU setting, the paper constructs modalities a, b, c where a→c and c→b align at width w while direct a→b needs width about w², showing that non-transitive compatibility and bridging are provable phenomena rather than anecdotal observations. A sympathetic reader would care because these invariants give a common substrate for comparing data types that are otherwise hard to compare.

What carries the argument

The projection-parameter sheaf assigns the same parameter space R^p to every vertex and edge of the fixed graph G, with identity restriction maps. Its 0-Laplacian energy is exactly Σ_{(u,v)} ||w_u − w_v||², the standard vector-valued graph Dirichlet energy, so the obstruction C_{a→b}(ε) is computed by sheaf-regularized regression without needing frames or transports. The other load-bearing object is the nested projection family: orthogonal linear maps ⊂ low-rank linear maps ⊂ bounded-width Lipschitz MLPs, all applied after whitening; the complexity index α in this family defines H. The spectral gap λ₂(G) enters through Poincaré's inequality and connects local fitting quality to global-map er

What would settle it

Take the paper's two-cluster sign-flip model on a graph with a known cut, fit scalar parameters +1/−1 per vertex, and compute the minimum variation energy among all perfect-fitting fields; the paper predicts exactly 4|E(V⁺,V⁻)|. If a brute-force search over perfect-fitting assignments finds a lower energy, Proposition 7 fails. Alternatively, instantiate the one-dimensional ReLU construction with dense noiseless samples; zero-error approximation of a Θ(w²)-breakpoint function by a width-w one-hidden-layer network would contradict Lemma 4.

Watch

Extended reading notes

Core claim

The central claim is that any directed modality pair (a→b) can be characterized by two complementary, computable invariants on a fixed modality-independent site G: projection hardness H_{a→b}(ε), the minimal complexity inside a nested Lipschitz-controlled projection family needed for one global map to reach error ε, and sheaf-Laplacian obstruction C_{a→b}(ε), the minimal variation energy of a locally fit parameter field needed for the same error. The paper establishes that C is literally the 0-Laplacian energy of a projection-parameter sheaf with identity restrictions, which equals the quadratic smoothness penalty in sheaf-regularized regression; that small obstruction plus a spectral gap im

Load-bearing premise

The load-bearing premise is that the single neighborhood graph G, built once from latent states, labels, or consensus, approximates the true semantic locality of the data; if G is wrong, the obstruction values and latent-level conclusions lose their meaning even though the formal inequalities still hold.

Editorial extensions

If this is right

If two modalities align through a bridge c with O(w) width in each stage but direct alignment demands Ω(w²), then compatibility cannot be assumed transitive; systems should search for pivots rather than force a direct map.
Obstruction C and hardness H form two orthogonal axes: a pair can be hard but glueable, or easy but non-gluable; diagnosing which one fails changes the remedy, more capacity versus better site or better local alignment.
Theorem 3 gives a quantitative bound: if locally varying fits are good and their parameter variation is small, the global mean parameter map is almost as good, with degradation controlled by the spectral gap λ₂(G).
Because the site, whitening, projection classes, and error metric are fixed, the numbers H and C become comparable across modality pairs and across datasets that share the same formal conventions.

Reading between the lines

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

Editorial extension: the framework suggests a practical diagnostic: compare H_{a→b} with H_{a→c}+H_{c→b}; a large gap would predict which pairs benefit from bridge modalities before training a full multiway aligner.
Editorial extension: the sign-flip example implies that semantically coherent groups that are internally consistent but flipped relative to each other will show up as high C concentrated on the cut between groups, which could be used to detect label inconsistencies or distinct semantic regimes.
Editorial extension: the paper leaves implicit that site construction itself could be optimized; given candidate site graphs, one could select G that minimizes overall obstruction across all pairs, turning the site-dependence from a limitation into a design parameter.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The framework is worth a serious look, but the paper's flagship example doesn't survive contact with its own definitions. The basic idea is good: put all modalities on a fixed graph site, then separate global expressivity (H) from local gluing failure (C). The projection-parameter sheaf with identity restrictions is a clean way to make obstruction a graph Dirichlet energy, and Theorem 3 (small obstruction implies near-global consistency when the loss is parameter-Lipschitz) is correct and standard. The spectral-gap control is exactly what you'd want. The paper is honestly written as a reference formalism, and most of the formal scaffolding is fine.

The problem is Theorem 6, the ReLU construction. Definition 5 requires whitened embeddings and a projection family with a fixed Lipschitz bound L. The construction sets z_a=x, z_c=g(x), z_b=h(g(x)), ignoring whitening entirely. And once you respect the fixed-L constraint, the Ω(w^2) breakpoint argument collapses. To get h∘g to have ~w^2 breakpoints, h's breakpoints have to lie in an interval of width O(L/w); otherwise each piece of g can only hit O(1) of them. But then h's output variation over that interval is O(1/w), so the target h∘g has range O(1/w). For any fixed tolerance ε, a constant map achieves error ε once w is large enough. So H_{a→b}(ε) is not Ω(w^2) under the paper's own normalization and Lipschitz constraint. The breakpoint counting in Lemma 5 is right as a statement about piecewise-linear composition, but it ignores the total-variation budget that a fixed L imposes. The abstract also claims synthetic calibrations that the body explicitly defers to future work, which is an overclaim.

I want to be fair: the framework itself may be salvageable. If you allow L to grow with w, the separation might hold, but then it's not a fixed family and the comparison is much less clean. The whitening issue might be fixable by constructing the examples in whitened coordinates, but that isn't done here. For a theory paper, the central existence proof should be solid, and this one isn't.

The paper deserves peer review because the conceptual contribution is novel and the flaws are specific and addressable, not because the current version is correct. I'd send it to a serious referee and ask for a corrected construction or a withdrawal of the non-transitivity claim. Cite it if you want the H/C framework as a reference point, but don't build on Theorem 6.

Referee Report

4 major / 5 minor

Summary. The paper develops a graph-sheaf formalism for cross-modal alignment on a fixed, modality-independent site. It defines two directed invariants: projection hardness H_{a→b}(ε), the minimum complexity in a Lipschitz-controlled family needed for a single global map to reach error ε on whitened embeddings, and sheaf-Laplacian obstruction C_{a→b}(ε), the minimum spatial variation of locally fitted projection parameters needed to reach the same error. The obstruction is implemented through a projection-parameter sheaf with identity restrictions, so its 0-Laplacian energy is exactly a vector-valued graph Dirichlet form. The paper proves a Poincaré-type bound linking obstruction energy to excess global-map error (Theorem 3), gives a cut-induced obstruction calculation (Proposition 7), and presents a ReLU construction claiming non-transitivity and a quadratic width separation between staged and direct alignment (Theorem 6). The abstract also claims controlled synthetic calibrations. I find the definitional framework and Theorem 3 sound, but the central non-transitivity theorem is not established under the paper's own normalization and tolerance definitions, and the calibration claim is absent from the manuscript.

Significance. If the results held, the framework would be a useful operational decomposition of alignment failure into a global expressivity axis and a local gluing axis, with an explicit spectral-gap control. The obstruction formalism is clean and directly computable, and the Poincaré bound (Theorem 3) and the cut-induced obstruction computation (Proposition 7) are correct and clearly presented. However, the paper contains no experiments despite the abstract's calibration claims, and the flagship quadratic separation is unsupported: the lower bound relies on exact breakpoint counting and ignores approximation error at fixed tolerance, which is incompatible with the definition of hardness. The non-transitivity claim is load-bearing for the paper's stated contribution, so the significance of the manuscript in its current form is substantially reduced.

major comments (4)

[§3.8, Theorem 6 and Definition 5] The lower bound H_{a→b}(ε)=Ω(w²) is not valid for the tolerance-based Definition 5. Under the spectral-norm constraints defining G_mlp_w (Eq. (12)), the sum of slope-change magnitudes of any width-w network is bounded by L. For h∘g to have Θ(w²) breakpoints, the breakpoints of h must lie in an interval of width O(1/w); hence h, and therefore h∘g, differs from an affine function by O(L/w) in sup norm. For any fixed ε>0, an affine map has error <ε for all sufficiently large w, so H_{a→b}(ε)=O(1). Breakpoint counting (Lemma 4) gives only an exact-representation lower bound and cannot support a positive gap at positive tolerance.
[§3.8, Eq. (9)] The construction does not apply the canonical whitening (Eq. (9)), although H is defined on whitened embeddings. The maps g and h are given between raw coordinates (or between arbitrary ̃z coordinates with no covariance check), but whitening rescales each modality by its empirical covariance, changing the Lipschitz constants of the stage maps by factors σ_a/σ_c and σ_c/σ_b. The O(w) upper bounds for H_{a→c} and H_{c→b} are therefore not established for the fixed-L family G_mlp_w. A concrete verification of all three hardness values in the whitened coordinate system is missing.
[Abstract vs §1 Scope] The abstract states that 'controlled synthetic calibrations recover the predicted hardness separation, cut-induced obstruction scaling, and sensitivity to the fixed site graph,' but the manuscript contains no experiments or calibration section. The Scope paragraph explicitly defers synthetic validation to future work. This discrepancy must be resolved: either remove the calibration claims from the abstract or include the supporting experiments.
[§3.8, Lemmas 4–5] Lemma 5 is only a proof sketch and does not construct g,h that satisfy the Lipschitz and whitening constraints required by G_mlp_w. Theorem 6 then rests on non-explicit 'appropriate choices'. Since the central separation depends on these maps, the existence claim needs a rigorous construction with explicit parameter assignments and verification of all constraints; as it stands, Theorem 6 is not proved.

minor comments (5)

[Theorem 3 proof and Appendix A] The proof says 'apply Theorem 2' but the referenced statement is Lemma 2; the same typo appears in Appendix A.
[§3.8] The text refers to 'Theorem 5' for the composition guarantee; this should be Lemma 5.
[Definition 5 and Eq. (12)] The nested family G_α is indexed by a single scalar α, but the ReLU construction treats width w as the only index and silently fixes depth and Lipschitz bound L. The ordering of α across the orthogonal, low-rank, and MLP classes should be stated explicitly to make H well-defined.
[Appendix B] The stagewise obstruction C_{a→c→b}(ε) is defined operationally via a two-parameter grid search over λ_1, λ_2. This is not a formal definition of the infimum and should be written as an explicit optimization over parameter fields, not just an algorithmic recipe.
[Figure 3] The right panel caption says 'obstruction despite perfect local fits' but no local fits are displayed; the figure should be made self-explanatory or the caption adjusted.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: H and C are explicit definitions, Theorem 3 is proved from a stated Lipschitz assumption, and the ReLU construction is a mathematical existence proof rather than a fitted prediction.

full rationale

The paper's derivation chain is definitional and theorem-based, not fitted or self-citational. Projection hardness (Definition 5) and sheaf-Laplacian obstruction (Definition 7) are explicit definitions, not derived predictions. The sheaf energy in Eq. (16) is deliberately set to the graph Dirichlet form (Remark 2), and the paper itself calls this a design choice ('This choice is deliberate for operationality'), so the exact match between the projection-parameter sheaf energy and the sheaf-regularized regression penalty is by construction rather than a discovered equivalence. Lemma 2 and Theorem 3 are proved from the standard graph Poincaré inequality plus an explicit parameter-Lipschitz assumption; the Lipschitz assumption is stated in Definition 8 and the proof is given in full, so no input is used as its own output. The ReLU non-transitivity construction (Lemmas 4–5, Theorem 6) is an existence construction: the modalities are defined by composing one-hidden-layer networks, and the O(w) vs Ω(w²) separation follows from the breakpoint-count argument, not from fitting a parameter to the claimed hardness values. No load-bearing self-citations or imported uniqueness theorems appear. A reviewer concern that the construction may not respect the whitening/Lipschitz conventions of Definition 5 and Eq. (9) would be a correctness or consistency objection to the construction, not a circularity: it does not involve the paper reducing a prediction to its own inputs. Similarly, the abstract's mention of synthetic calibrations conflicts with the Scope statement that empirical validation is deferred, but an unsubstantiated claim is not a circular derivation.

Assumptions & free parameters 1 free parameters · 7 assumptions · 1 invented entities

Central derivation rests on standard sheaf and graph-spectral facts plus a chain of domain assumptions: existence of a shared site that approximates semantic locality, identity-restricted parameter sheaf, Lipschitz-continuous loss, and an unproven construction of ReLU networks realizing Ω(w²) pieces.

free parameters (1)

site graph G (and k, σ in synthetic construction) = not specified (k, σ left open)
Definition 1 requires a fixed graph on sample indices; all obstruction and spectral values depend on it. In non-synthetic settings no principled construction is supplied, so G is a hand-chosen input that determines C_{a→b} and λ2(G).

assumptions (7)

standard math Cellular sheaf Laplacian and Dirichlet form formalism (Definitions 3-4)
Used throughout §3.1; standard applied sheaf theory.
standard math Graph Poincaré inequality with constant 1/λ2(G) for connected G
Lemma 2 and Appendix A; standard linear algebra.
domain assumption Modality-independent site requirement: one fixed graph G for all modalities
Definition 1; required for comparability, but restricts real settings where each modality induces its own geometry.
domain assumption Projection-parameter sheaf uses identity restrictions (constant sheaf)
§3.5 and §5; makes energy equal to vector Dirichlet form; explicitly left as a limitation.
domain assumption Per-sample loss is L_w-Lipschitz in projection parameters
Definition 8; required for Theorem 3; the paper states it is a modeling assumption.
domain assumption G approximates latent semantic locality; latent-level interpretation
Definition 11 and Proposition 8 are informal; low H/C only meaningful as latent-level statements under this approximation.
ad hoc to paper Existence of width-w ReLU g,h with h∘g having Θ(w²) breakpoints (Lemma 5/Thm 6)
The construction is only sketched ('for appropriate choices') and is needed for the quadratic separation.

invented entities (1)

Projection-parameter sheaf P (identity restrictions)
purpose: Defines the variation energy E_P(w)=Σ||w_u-w_v||² used in C_{a→b}(ε).
Definition 6; a mathematical bookkeeping device whose energy matches the smoothness penalty by construction, with no empirical handle outside the framework.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Sheaf-Laplacian Obstruction and Projection Hardness for Cross-Modal Compatibility on a Modality-Independent Site." pith.science (2026). https://pith.science/paper/J6OUKN24

@misc{pith2026260407632,
  author       = {Pith},
  title        = {Pith review of: Sheaf-Laplacian Obstruction and Projection Hardness for Cross-Modal Compatibility on a Modality-Independent Site},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J6OUKN24}},
  note         = {Machine review of arXiv:2604.07632}
}

abstract

Cross-modal representations vary in how easily they can be aligned, and compatibility is generally non-transitive: two modalities may align through an intermediate modality at lower complexity than through a direct map. We introduce a reference formalism that evaluates all modalities on a fixed neighborhood site and defines two directed invariants. Projection hardness \(H_{a\to b}(\varepsilon)\) is the minimum complexity within a nested Lipschitz-controlled family required to reach error \(\varepsilon\). For a declared local projection family, sheaf-Laplacian obstruction \(C_{a\to b}(\varepsilon)\) is the minimum variation of locally fitted projection parameters required to reach the same error. Under identity restrictions, obstruction is the graph Dirichlet energy of a vector-valued parameter field; the sheaf formulation identifies zero energy with successful gluing and extends to edge-dependent transports and heterogeneous parameter spaces. We relate obstruction to the site spectral gap and excess global-map error, and construct ReLU examples showing non-transitive compatibility and a quadratic separation between staged and direct width. Controlled synthetic calibrations recover the predicted hardness separation, cut-induced obstruction scaling, and sensitivity to the fixed site graph.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references · 1 linked inside Pith

[1]

IEEE Transactions on Pattern Analysis and Machine Intelligence41(2), 423–443 (2018)

Baltrusaitis, T., Ahuja, C., Morency, L.-P.: Multimodal machine learning: A survey and taxonomy. IEEE Transactions on Pattern Analysis and Machine Intelligence41(2), 423–443 (2018)

2018
[2]

Biometrika28(3/4), 321–377 (1936) https://doi.org/10.1093/biomet/28.3-4.321

Hotelling, H.: Relations between two sets of variates. Biometrika28(3/4), 321–377 (1936) https://doi.org/10.1093/biomet/28.3-4.321

doi:10.1093/biomet/28.3-4.321 1936
[3]

Inter- national Journal of Neural Systems10(05), 365–377 (2000) https://doi.org/10

Lai, P.L., Fyfe, C.: Kernel and nonlinear canonical correlation analysis. Inter- national Journal of Neural Systems10(05), 365–377 (2000) https://doi.org/10. 1142/S012906570000034X

2000
[4]

In: Proceedings of the 38th International Conference on Machine Learning

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning transferable visual models from natural language supervision. In: Proceedings of the 38th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 139, pp. 8748–8763 (2021)

2021
[5]

PhD thesis, University of Pennsylvania (2014)

Curry, J.M.: Sheaves, cosheaves and applications. PhD thesis, University of Pennsylvania (2014)

2014
[6]

CreateSpace Independent Publishing Platform, Charleston, SC (2014)

Ghrist, R.: Elementary Applied Topology, 1st edn. CreateSpace Independent Publishing Platform, Charleston, SC (2014)

2014
[7]

In: NeurIPS Workshop on Topo- logical Data Analysis and Beyond (2020)

Hansen, J., Gebhart, T.: Sheaf neural networks. In: NeurIPS Workshop on Topo- logical Data Analysis and Beyond (2020). Key corresponds to the generalization of GCNs via Sheaf Laplacians

2020
[8]

In: Advances in Neural Information Processing Systems, vol

Bodnar, C., Di Giovanni, F., Chamberlain, B., Li´ o, P., Bronstein, M.: Neural sheaf diffusion: A topological perspective on heterophily and oversmoothing in gnns. In: Advances in Neural Information Processing Systems, vol. 35, pp. 18527–18541 (2022)

2022

Show all 18 references

[9]

New Journal of Physics25(9), 093013 (2023) https://doi.org/ 10.1088/1367-2630/acf33c

Calmon, L., Schaub, M.T., Bianconi, G.: Dirac signal processing of higher-order topological signals. New Journal of Physics25(9), 093013 (2023) https://doi.org/ 10.1088/1367-2630/acf33c

2023 doi
[10]

Machine Learning79(1), 151–175 (2010) https://doi.org/10.1007/s10994-009-5152-4

Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., Vaughan, J.W.: A theory of learning from different domains. Machine Learning79(1), 151–175 (2010) https://doi.org/10.1007/s10994-009-5152-4

2010 doi
[11]

In: Proceedings of the 22nd Annual Conference on Learning Theory (COLT)

Mansour, Y., Mohri, M., Rostamizadeh, A.: Domain adaptation: Learning bounds and algorithms. In: Proceedings of the 22nd Annual Conference on Learning Theory (COLT). Omnipress, Montreal, Canada (2009)

2009
[12]

The Annals of Statistics49(1), 100–128 (2021)

Cai, T.T., Wei, H.: Transfer learning for nonparametric classification: Minimax 20 rate and adaptive classifier. The Annals of Statistics49(1), 100–128 (2021)

2021
[13]

In: Advances in Neural Information Processing Systems (NIPS), vol

Perrot, M., Courty, N., Flamary, R., Habrard, A.: Mapping estimation for dis- crete optimal transport. In: Advances in Neural Information Processing Systems (NIPS), vol. 29, pp. 4197–4205 (2016)

2016
[14]

arXiv preprint arXiv:2209.03430 (2022)

Liang, P.P., Zadeh, A., Morency, L.-P.: Foundations and trends in multimodal machine learning: Principles, challenges, and open questions. arXiv preprint arXiv:2209.03430 (2022)

2022 arXiv
[15]

In: Proceedings of the 41st International Conference on Machine Learning (ICML) (2024)

Wu, S., Fei, H., Qu, L., Ji, W., Chua, T.-S.: NExT-GPT: Any-to-any multimodal LLM. In: Proceedings of the 41st International Conference on Machine Learning (ICML) (2024)

2024
[16]

In: Pro- ceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pp

Firat, O., Sankaran, B., Al-Onaizan, Y., Yarman-Vural, F.T., Cho, K.: Zero- resource translation with multi-lingual neural machine translation. In: Pro- ceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pp. 268–277 (2016)

2016
[17]

In: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp

Gu, J., Hassan, H., Devlin, J., Li, V.O.: Universal neural machine translation for extremely low resource languages. In: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 344–354 (2018)

2018
[18]

Graduate Studies in Mathematics, vol

Evans, L.C.: Partial Differential Equations. Graduate Studies in Mathematics, vol. 19. American Mathematical Society, Providence, RI (1998) 21

1998

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

IEEE Transactions on Pattern Analysis and Machine Intelligence41(2), 423–443 (2018)

Baltrusaitis, T., Ahuja, C., Morency, L.-P.: Multimodal machine learning: A survey and taxonomy. IEEE Transactions on Pattern Analysis and Machine Intelligence41(2), 423–443 (2018)

2018

[2] [2]

Biometrika28(3/4), 321–377 (1936) https://doi.org/10.1093/biomet/28.3-4.321

Hotelling, H.: Relations between two sets of variates. Biometrika28(3/4), 321–377 (1936) https://doi.org/10.1093/biomet/28.3-4.321

doi:10.1093/biomet/28.3-4.321 1936

[3] [3]

Inter- national Journal of Neural Systems10(05), 365–377 (2000) https://doi.org/10

Lai, P.L., Fyfe, C.: Kernel and nonlinear canonical correlation analysis. Inter- national Journal of Neural Systems10(05), 365–377 (2000) https://doi.org/10. 1142/S012906570000034X

2000

[4] [4]

In: Proceedings of the 38th International Conference on Machine Learning

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning transferable visual models from natural language supervision. In: Proceedings of the 38th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 139, pp. 8748–8763 (2021)

2021

[5] [5]

PhD thesis, University of Pennsylvania (2014)

Curry, J.M.: Sheaves, cosheaves and applications. PhD thesis, University of Pennsylvania (2014)

2014

[6] [6]

CreateSpace Independent Publishing Platform, Charleston, SC (2014)

Ghrist, R.: Elementary Applied Topology, 1st edn. CreateSpace Independent Publishing Platform, Charleston, SC (2014)

2014

[7] [7]

In: NeurIPS Workshop on Topo- logical Data Analysis and Beyond (2020)

Hansen, J., Gebhart, T.: Sheaf neural networks. In: NeurIPS Workshop on Topo- logical Data Analysis and Beyond (2020). Key corresponds to the generalization of GCNs via Sheaf Laplacians

2020

[8] [8]

In: Advances in Neural Information Processing Systems, vol

Bodnar, C., Di Giovanni, F., Chamberlain, B., Li´ o, P., Bronstein, M.: Neural sheaf diffusion: A topological perspective on heterophily and oversmoothing in gnns. In: Advances in Neural Information Processing Systems, vol. 35, pp. 18527–18541 (2022)

2022