Pith. sign in

REVIEW 2 major objections 4 minor 74 references

A Diagrammatic Approach to Improve Computational Efficiency in Group Equivariant Neural Networks

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read For four symmetry groups, every equivariant weight matrix can be factored through a planar diagram, cutting forward-pass cost from O(n^(l+k)) to O(n^k) or better.

desk verdict The categorical reformulation is neat and the O(n)/Sp(n)/SO(n) extensions are new, but the headline Big-O claims hold only under a zero-cost output-materialization assumption that won't survive dense tensor practice. read the letter →

arxiv 2412.10837 v1 pith:2VFUKNQC submitted 2024-12-14 cs.LG math.COmath.RTstat.ML

Edward Pearce-Crump , William J. Knottenbelt This is my paper · ORCID
classification cs.LGmath.COmath.RTstat.ML
keywords deeplearningtheoryequivariantneuralnetworksweightmatricesdiagrammaticmethodsmonoidalcategoriespartitiondiagramstensorpowerrepresentationstimecomplexity
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

Group equivariant neural networks gain their inductive bias by building known symmetries into every layer, but when the layers are high-order tensor powers of R^n, a naive forward pass through an equivariant weight matrix costs O(n^(l+k)) operations. This paper establishes that for the symmetric, orthogonal, special orthogonal, and symplectic groups, every such weight matrix can be factored through a diagrammatic normal form, reducing the forward pass to O(n^k) or better. The factorisation first expresses each weight matrix as a linear combination of set-partition diagrams via monoidal functors, then uses the string-like flexibility of those diagrams to split each computation into permutations, tensor contractions, identity transfers, and copies. If the cost analysis holds, high-order equivariant layers stop being prohibitively expensive and can be used in practical networks.

What carries the argument

The central object is the set-partition diagram: a two-row diagram whose connected components record which tensor indices are identified. The new twist is the algorithmically planar diagram, a set-partition diagram whose bottom-row components sit at the far right in decreasing size, whose top-row components sit at the far left, and in which no components cross. Every diagram for the four groups factors as a permutation, an algorithmically planar diagram, and a permutation; because the planar diagram is a tensor product of the smallest possible blocks, applying the corresponding monoidal functor converts it into a Kronecker product of tiny equivariant matrices. The left and right permutations are executed as basis reorderings, and the Kronecker product is executed right-to-left as contractions, identity transfers, and copies.

What would settle it

Fix k=2 and l=10, take an O(n)-equivariant map given by a Brauer diagram with one bottom-row pair, and count operations in a dense implementation: the diagrammatic algorithm performs O(n) arithmetic (one contraction over n terms) but must write $n^{10}$ output coefficients. If memory writes are counted, the total cost scales as $n^{10}$, not O(n^(k-1)); this is a direct observation that separates the arithmetic claim from the full forward-pass claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the expensive equivariant weight matrices of tensor-power networks are not monolithic: for the symmetric, orthogonal, special orthogonal, and symplectic groups, each matrix is the image of a set-partition diagram under a full monoidal functor, and each diagram can be factored into a permutation, an algorithmically planar diagram, and another permutation. The planar diagram then decomposes as a tensor product of irreducible single-block diagrams, so the corresponding matrix is a Kronecker product of small matrices that act as tensor contractions, identity transfer operations, and copy operations. Applying these components right-to-left computes the full matrix-vector product; the paper's complexity analysis gives O(n^k) for the symmetric group, O(n^(k-1)) for the orthogonal and symplectic groups, and O(n^(k-(n-s))(n! + n^(s-1))) for the free-vertex diagrams of the special orthogonal group, compared with O(n^(l+k)) for direct multiplication.

Load-bearing premise

The paper's Remark 37 says memory operations like permuting basis vectors and copying coefficients cost nothing; if that assumption is removed, producing a dense n^l-dimensional output already costs O(n^l) operations, so the advertised O(n^(k-1)) exponents are not the full runtime.

Editorial extensions

If this is right

  • For S_n-equivariant layers, the forward pass through any weight matrix costs O(n^k) in the paper's arithmetic model, replacing the naive O(n^(l+k)).
  • For O(n) and Sp(n), the cost becomes O(n^(k-1)); for SO(n), Brauer-type elements match O(n) and free-vertex diagrams cost O(n^(k-(n-s))(n! + n^(s-1))).
  • Because each spanning-set element can be multiplied separately and combined linearly, a full weight matrix can be applied by running the fast multiplication in parallel over its diagram components.
  • For the orthogonal, symplectic, and special-orthogonal groups, the middle transfer factor reduces to the identity, so the remaining work is tensor contractions and copy or index operations.
  • The factorisation applies to every equivariant weight matrix in these networks, not only special cases, since every matrix is a linear combination of spanning-set elements.

Reading between the lines

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

  • The factoring recipe should extend to any group whose equivariant maps admit a set-partition-diagram description, such as alternating groups or direct products, as long as an algorithmically planar normal form exists.
  • The complexity claims are arithmetic-centric: in a dense explicit representation, writing the n^l output coefficients costs O(n^l) memory operations, so the advertised O(n^(k-1)) exponents for O(n) and Sp(n) best describe implicit-output or memory-free models.
  • A practical implementation could compile PlanarMult into standard tensor-contraction kernels; if the contraction order follows the right-to-left block order, no custom fused kernels should be needed to realise the complexity.
  • A sharper test of the central claim is to hold k fixed, raise l, and measure arithmetic operations for increasing n: the paper predicts the arithmetic count should not grow with l, which is a stronger statement than wall-clock time because wall-clock includes the free memory operations.
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 / 4 minor

Summary. The paper proposes a diagrammatic factorization algorithm for applying equivariant weight matrices between tensor power spaces for four groups: the symmetric, orthogonal, special orthogonal, and symplectic groups. It introduces strict monoidal categories of partition diagrams and shows that the known equivariant map characterizations arise from full monoidal functors. The main contribution, Algorithm 1 (MatrixMult), factors an equivariant weight matrix as a permutation, an "algorithmically planar" diagram, and another permutation, then applies the planar part through a sequence of contractions, transfers, and copies. The paper claims that this reduces the time complexity from the naive O(n^{l+k}) to O(n^k) for S_n, O(n^{k-1}) for O(n) and Sp(n), and O(n^{k-(n-s)}(n! + n^{s-1})) for SO(n), and that in some cases the computation is "effectively free."

Significance. If the stated time-complexity claims were correct, the paper would be a substantial practical advance for high-order tensor-power equivariant networks. The category-theoretic reformulation is elegant, the functoriality proof for S_n is self-contained and checks out, and the worked examples in Section 5.2 are consistent with the index manipulations. The decomposition into contractions, transfers, and copies is a genuinely useful way to organize the computation. However, the headline complexity claims conflate arithmetic-operation counts with wall-clock time and ignore the cost of materializing the dense output tensor. The corrected complexity still gives a meaningful speedup over naive multiplication, but the specific bounds in Section 5.2 and the abstract's "Big-O time complexity" claim need revision. The paper provides no code or machine-checked implementation, so the practical claims rest entirely on the described algorithm.

major comments (2)
  1. [Section 5.2, Remark 37] The complexity analysis assumes that "memory operations, such as permuting basis vectors and making copies of coefficients, have no cost." This makes the stated bounds arithmetic-operation counts for an implicitly represented tensor, not time complexities for applying a dense layer. The output of the layer is an element of (R^n)^l, and any concrete dense representation must write n^l coefficients. For example, take k=1, l=2 and a (1,2)-partition diagram with one bottom singleton and one top pair: Step 1 costs O(n) and Step 3 is declared free, but the output is a dense n^2-vector, so the stated O(n^k)=O(n) bound is false. The same issue affects the O(n^{k-1}) bounds for O(n) and Sp(n) and the SO(n) bound in equation (169). The abstract's "Big-O time complexity" claim and the concluding "effectively free" statement therefore need revision; a corrected statement that includes an Omega(n^l) output term still preserves a substantial speedup over O(n^{l+k}), but the specific exponents in Section 5.2 are not time complexities as written.
  2. [Section 5.2.1, Steps 2-3 and time-complexity paragraph] The claim that transfer and copy operations have "no cost" hides the fact that these steps are what create the output tensor's size. In the S_n algorithm, Step 3 takes the contracted vector and produces a vector with n^l coefficients by repeating indices; this is an expansion of the representation, not a no-op. The "best case" statement that "the computation is effectively free" when b=0 is untenable whenever l>0, because the output still has n^l entries. The authors should either adopt an explicit sparse or implicit tensor representation and explain how it is consumed by subsequent layers, or add the materialization cost to all reported bounds.
minor comments (4)
  1. [Section 5.2, Factor] The Factor procedure is described informally as "drag and bend the strings" rather than as a fully specified algorithm. Please state explicitly whether the factorization is precomputed once per weight matrix (with its cost charged separately) or is part of each forward pass; the zero-cost assumption in Remark 37 makes this distinction important.
  2. [Section 5.2.2, equation (124)] The symbol v_L is used for the coefficient of the vector entering Step 3, although that vector was called w in the surrounding text; please harmonize the notation.
  3. [Section 5.2.4, equation (169)] The expression contains an n! factor. Since n is the dimension parameter appearing in the Big-O statements, please clarify whether n is treated as fixed or growing, because n! changes the asymptotic class in a way that is not obvious from the notation O(n^{k-(n-s)}(n! + n^{s-1})).
  4. [Section 5.2.1, equations (115)-(116)] The text describes the contraction step as involving "n terms to multiply," but equation (98) is a summation; please clarify the unit-cost convention for multiplications versus additions or rephrase to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the algorithm implements previously characterized equivariant maps under an explicit arithmetic-cost model; self-citations point to published proofs and are not the claim being derived.

full rationale

The derivation chain is self-contained in the relevant sense. Theorems 5, 7, 9, and 11 supply spanning sets and bases for the equivariant maps; these are cited external results (Godfrey et al. 2023 for S_n; Pearce-Crump 2023a for O(n), Sp(n), and SO(n)), and the algorithm applies those maps rather than presupposing its own speedup. The category-theoretic functors in Section 4 are restatements of those surjections, not conclusions derived from the algorithm. The S_n algorithm is explicitly credited to Godfrey et al. (2023, Appendix C), so the acknowledged overlap is external provenance rather than circularity. The only substantial weakness is Remark 37, which declares memory operations costless; under that declared convention, the Big-O counts in Section 5.2 are arithmetic-operation counts, and if one insists on a dense output representation, an Omega(n^l) materialization term is omitted. That is a modeling assumption or correctness limitation, not a circular step, because it does not make any output equal to an input by construction. No fitted parameter is renamed as a prediction, no uniqueness claim is imported to forbid alternatives, and no central claim reduces to a self-citation chain.

Assumptions & free parameters 0 free parameters · 6 assumptions · 1 invented entities

The paper introduces no free parameters and no physical entities. The key invented construct is the algorithmically planar diagram. The central mathematical assumptions are the cited characterizations of equivariant maps and the zero-cost memory model, which is the main caveat to the complexity claims.

assumptions (6)
  • domain assumption The characterization of Hom_Sn as the span of partition diagram matrices (Theorem 5, cited to Godfrey et al. 2023).
    The paper cites this as a known result and uses it as the foundation for the symmetric group algorithm.
  • domain assumption The characterization of Hom_O(n) by Brauer diagrams (Theorem 7, cited to Pearce-Crump 2023a).
    Used for the orthogonal group algorithm.
  • domain assumption The characterization of Hom_Sp(n) by Brauer diagrams with symplectic form entries (Theorem 9, cited to Pearce-Crump 2023a).
    Used for the symplectic group algorithm.
  • domain assumption The characterization of Hom_SO(n) by Brauer diagrams and (l+k)\n diagrams (Theorem 11, cited to Pearce-Crump 2023a).
    Used for the special orthogonal group algorithm.
  • ad hoc to paper Memory operations (copying, permuting) have zero cost in the complexity model (Remark 37).
    This assumption underlies the stated time complexity bounds and is not standard for dense tensor implementations.
  • domain assumption Set partition diagrams can be freely deformed as string diagrams without changing the represented linear map (monoidal category string diagram equivalence).
    Used in Factor to reorder vertices and ensure planarity; relies on the functoriality theorems.
invented entities (1)
  • Algorithmically planar set partition diagram
    purpose: A normal form for diagrams that allows factorization into a tensor product of simple diagrams, enabling the fast multiplication algorithm.
    This is a new mathematical definition introduced in Section 5.1; it has no empirical handle outside the paper but is internally well-defined.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Diagrammatic Approach to Improve Computational Efficiency in Group Equivariant Neural Networks." pith.science (2026). https://pith.science/paper/2VFUKNQC

@misc{pith2026241210837,
  author       = {Pith},
  title        = {Pith review of: A Diagrammatic Approach to Improve Computational Efficiency in Group Equivariant Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2VFUKNQC}},
  note         = {Machine review of arXiv:2412.10837}
}
abstract

Group equivariant neural networks are growing in importance owing to their ability to generalise well in applications where the data has known underlying symmetries. Recent characterisations of a class of these networks that use high-order tensor power spaces as their layers suggest that they have significant potential; however, their implementation remains challenging owing to the prohibitively expensive nature of the computations that are involved. In this work, we present a fast matrix multiplication algorithm for any equivariant weight matrix that maps between tensor power layer spaces in these networks for four groups: the symmetric, orthogonal, special orthogonal, and symplectic groups. We obtain this algorithm by developing a diagrammatic framework based on category theory that enables us to not only express each weight matrix as a linear combination of diagrams but also makes it possible for us to use these diagrams to factor the original computation into a series of steps that are optimal. We show that this algorithm improves the Big-$O$ time complexity exponentially in comparison to a na\"{i}ve matrix multiplication.

Figures

Figures reproduced from arXiv: 2412.10837 by the authors.

Figure 1
Figure 1. We use the string-like property of (k, l)–partition diagrams to Factor them as a composition of a permutation in Sk, an algorithmically planar (k, l)–partition diagram, and a permutation in Sl . Here, k = 5 and l = 4. Remark 37 For each group G(n), we also analyse the time complexity of the implementa￾tion of MatrixMult and compare it with the time complexity of the naive implementation of matrix multiplication. Not… view at source ↗
Figure 2
Figure 2. The decomposition of the algorithmically planar ( [PITH_FULL_IMAGE:figures/full_fig_p029_2.png] view at source ↗
Figure 3
Figure 3. We show how matrix multiplication is implemented i [PITH_FULL_IMAGE:figures/full_fig_p030_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: We use the string-like aspect of (k, l)–Brauer diagrams to Factor them as a com￾position of a permutation in Sk, an algorithmically planar (k, l)–Brauer diagram, and a permutation in Sl . Here k = l = 5. Given how Factor constructs the planar (k, l)–Brauer diagram dβ, …
Figure 5
Figure 5. Figure 5: The tensor product decomposition of the planar (5 [PITH_FULL_IMAGE:figures/full_fig_p035_5.png]
Figure 6
Figure 6. Figure 6: We show how matrix multiplication is implemented i [PITH_FULL_IMAGE:figures/full_fig_p037_6.png]
Figure 7
Figure 7. Figure 7: We use the string-like aspect of (l + k)\n–diagrams to Factor them as a compo￾sition of a permutation in Sk, an algorithmically planar (l + k)\n–diagram, and a permutation in Sl . Here, k = 5 and l = 4. Time Complexity The time complexity is exactly the same as for the…
Figure 8
Figure 8. Figure 8: The tensor product decomposition of the algorithm [PITH_FULL_IMAGE:figures/full_fig_p043_8.png]
Figure 9
Figure 9. Figure 9: We show how matrix multiplication is implemented i [PITH_FULL_IMAGE:figures/full_fig_p045_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 66 canonical work pages

  1. [1]

    V. Abbott. Neural Circuit Diagrams: Robust Diagrams for the Communication, Implementation, and Analysis of Deep Learning Architectures . Transactions on Machine Learning Research, 2024

    work page 2024

  2. [2]

    Benkart and T

    G. Benkart and T. Halverson. Partition algebras P_k(n) with 2k>n and the fundamental theorems of invariant theory for the symmetric group S_n . Journal of the London Mathematical Society, 99 0 (2): 0 194--224, 2019 a

    work page 2019

  3. [3]

    Benkart and T

    G. Benkart and T. Halverson. Partition A lgebras and the I nvariant T heory of the S ymmetric G roup. In Recent Trends in Algebraic Combinatorics, volume 16 of Association for Women in Mathematics Series, pages 1--41. Springer, 2019 b

    work page 2019

  4. [4]

    Benkart, T

    G. Benkart, T. Halverson, and N. Harman. Dimensions of irreducible modules for partition algebras and tensor power multiplicities for symmetric and alternating groups. Journal of Algebraic Combinatorics. An International Journal, 46 0 (1): 0 77--108, 2017

    work page 2017

  5. [5]

    Bloem-Reddy and Y

    B. Bloem-Reddy and Y. W. Teh. Probabilistic Symmetries and Invariant Neural Networks . Journal of Machine Learning Research, 21: 0 1--61, 2020

    work page 2020

  6. [6]

    Bogatskiy, B

    A. Bogatskiy, B. Anderson, J. Offermann, M. Roussi, D. Miller, and R. Kondor. Lorentz Group Equivariant Neural Network for Particle Physics . In International Conference on Machine Learning, pages 992--1002. PMLR, 2020

    work page 2020

  7. [7]

    R. Brauer. On A lgebras W hich A re C onnected with the S emisimple C ontinuous G roups. Annals of Mathematics, 38: 0 857--872, 1937

    work page 1937

  8. [8]

    W. P. Brown. An A lgebra R elated to the O rthogonal G roup . Michigan Mathematical Journal, 3 0 (1): 0 1 -- 22, 1955

    work page 1955

Show all 74 references
  1. [9]

    W. P. Brown. The S emisimplicity of the B rauer A lgebra. Annals of Mathematics, 63 0 (2): 0 324--335, 1956

    1956

  2. [10]

    Chatzipantazis, S

    E. Chatzipantazis, S. Pertigkiozoglou, E. Dobriban, and K. Daniilidis. SE(3)-Equivariant Attention Networks for Shape Reconstruction in Function Space . In The Eleventh International Conference on Learning Representations, 2023

    2023

  3. [11]

    Cho and B

    K. Cho and B. Jacobs. Disintegration and Bayesian inversion via string diagrams . Mathematical Structures in Computer Science, 29 0 (7): 0 938--971, 2019

    2019

  4. [12]

    Coecke and A

    B. Coecke and A. Kissinger. Picturing Quantum Processes: A First Course in Quantum Theory and Diagrammatic Reasoning . Cambridge University Press, 2017

    2017

  5. [13]

    Cohen and M

    T. Cohen and M. Welling. Group Equivariant Convolutional Networks . In Proceedings of The 33rd International Conference on Machine Learning, volume 48 of Proceedings of Machine Learning Research, pages 2990--2999, New York, USA, 20--22 Jun 2016. PMLR

    2016

  6. [14]

    Cohen and M

    T. Cohen and M. Welling. Steerable CNN s. In International Conference on Learning Representations, 2017

    2017

  7. [15]

    Cohen, M

    T. Cohen, M. Geiger, J. Köhler, and M. Welling. Spherical CNN s. In International Conference on Learning Representations, 2018

    2018

  8. [16]

    Cohen, M

    T. Cohen, M. Geiger, and M. Weiler. A General Theory of Equivariant CNNs on Homogeneous Spaces . In Advances in Neural Information Processing Systems, volume 32, 2019

    2019

  9. [17]

    J. Comes. Jellyfish P artition C ategories. Algebras and Representation Theory, 23: 0 327--347, 2020

    2020

  10. [18]

    R. Cornish. Stochastic Neural Network Symmetrisation in Markov Categories , 2024. arXiv:2406.11814

    2024 arXiv

  11. [19]

    Cruttwell, B

    G. Cruttwell, B. Gavranović, N. Ghani, P. Wilson, and F. Zanasi. Categorical Foundations of Gradient-Based Learning . In I. Sergey, editor, Programming Languages and Systems, volume 13240 of Lecture Notes in Computer Science. Springer, Cham, 2022

    2022

  12. [20]

    de Haan, T

    P. de Haan, T. S. Cohen, and M. Welling. Natural Graph Networks . In Advances in Neural Information Processing Systems, volume 33, pages 3636--3646, 2020

    2020

  13. [21]

    Duncan, A

    R. Duncan, A. Kissinger, S. Perdrix, and J. van de Wetering. Graph-theoretic Simplification of Quantum Circuits with the ZX-calculus . Quantum, 4, June 2020

    2020

  14. [22]

    Esteves, C

    C. Esteves, C. Allen-Blanchette, A. Makadia, and K. Daniilidis. Learning SO(3) Equivariant Representations with Spherical CNNs . In Proceedings of the European Conference on Computer Vision (ECCV), pages 52--68, 2018

    2018

  15. [23]

    Finzi, M

    M. Finzi, M. Welling, and A. G. Wilson. A P ractical M ethod for C onstructing E quivariant M ultilayer P erceptrons for A rbitrary M atrix G roups. In Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research,...

    2021

  16. [24]

    T. Fritz. A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics . Advances in Mathematics, 370: 0 107239, 2020

    2020

  17. [25]

    Gavranovi\' c , P

    B. Gavranovi\' c , P. Lessard, A. J. Dudzik, T. Von Glehn, J. a. G. Madeira Ara\' u jo, and P. Veli c kovi\' c . Position: Categorical Deep Learning is an Algebraic Theory of All Architectures . In Proceedings of the 41st International Conference on Machine Learning, volume 23...

    2024

  18. [26]

    Godfrey, M

    C. Godfrey, M. G. Rawson, D. Brown, and H. Kvinge. F ast computation of permutation equivariant layers with the partition algebra. In ICLR 2023 Workshop on Physics for Machine Learning, 2023

    2023

  19. [27]

    Gordon, D

    J. Gordon, D. Lopez-Paz, M. Baroni, and D. Bouchacourt. Permutation Equivariant Models for Compositional Generalization in Language . In International Conference on Learning Representations, 2020

    2020

  20. [28]

    C. Grood. B rauer A lgebras and C entralizer A lgebras for SO(2n,C) . Journal of Algebra, 222 0 (2): 0 678--707, 1999

    1999

  21. [29]

    Guttenberg, N

    N. Guttenberg, N. Virgo, O. Witkowski, H. Aoki, and R. Kanai. Permutation-equivariant neural networks applied to dynamics prediction, 2016. arXiv:1612.04530

    2016 arXiv

  22. [30]

    Halverson and A

    T. Halverson and A. Ram. Gems from the W ork of G eorgia B enkart. Notices of the American Mathematical Society, 69 0 (3): 0 375--384, 2022

    2022

  23. [31]

    Hanlon and D

    P. Hanlon and D. Wales. On the D ecomposition of B rauer's C entralizer A lgebras. Journal of Algebra, 121 0 (2): 0 409--445, 1989

    1989

  24. [32]

    J. S. Hartford, D. R. Graham, K. Leyton-Brown, and S. Ravanbakhsh. Deep M odels of I nteractions A cross S ets . In Proceedings of the 35th International Conference on Machine Learning , pages 1914--1923. PMLR , 2018

    1914

  25. [33]

    Heunen and J

    C. Heunen and J. Vicary. Categories for Quantum Theory: An Introduction. Oxford University Press, 2020

    2020

  26. [34]

    M. Hu. Presentations of D iagram C ategories. The PUMP Journal of Undergraduate Research, 3: 0 1--25, 2019

    2019

  27. [35]

    V. F. R. Jones. The P otts model and the symmetric group. In Subfactors: Proceedings of the T aniguchi Symposium on Operator Algebras ( K yuzeso, 1993) , pages 259--267. World Scientific, 1994

    1993

  28. [36]

    Jumper, R

    J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A. Žídek, A. Potapenko, A. Bridgland, C. Meyer, S. A. A. Kohl, A. J. Ballard, A. Cowie, B. Romera-Paredes, S. Nikolov, R. Jain, J. Adler, T. Back, S. Petersen, D. Reiman, E. C...

    2021

  29. [37]

    Khatri, T

    N. Khatri, T. Laakkonen, J. Liu, and V. Wang-Maścianica. On the Anatomy of Attention , 2024. arXiv:2407.02423

    2024 arXiv

  30. [38]

    Kissinger

    A. Kissinger. Pictures of processes: Automated graph rewriting for monoidal categories and applications to quantum computing, 2012. arXiv:1203.0202

    2012 arXiv

  31. [39]

    J. Kock. F robenius A lgebras and 2-D T opological Q uantum F ield T heories . London Mathematical Society Student Texts. Cambridge University Press, 2003

    2003

  32. [40]

    Kondor and S

    R. Kondor and S. Trivedi. On the Generalization of Equivariance and Convolution in Neural Networks to the Action of Compact Groups . In Proceedings of the 35th International Conference on Machine Learning, volume 80 of Proceedings of Machine Learning Research, pages 2747--2755...

    2018

  33. [41]

    Kondor, Z

    R. Kondor, Z. Lin, and S. Trivedi. Clebsch G ordan N ets: a F ully F ourier S pace S pherical C onvolutional N eural N etwork. In Advances in Neural Information Processing Systems, volume 31, 2018

    2018

  34. [42]

    A. D. Lauda and J. Sussan. A n I nvitation to C ategorification. Notices of the American Mathematical Society, 69 0 (1): 0 11--21, 2022

    2022

  35. [43]

    G. I. Lehrer and R. B. Zhang. The B rauer category and invariant theory. Journal of the European Mathematical Society, 17: 0 2311--2351, 2012

    2012

  36. [44]

    G. I. Lehrer and R. B. Zhang. I nvariants of the special orthogonal group and an enhanced B rauer category. L’Enseignement Mathématique, 63: 0 181--200, 2018

    2018

  37. [45]

    Leinster

    T. Leinster. Basic Category Theory. Cambridge University Press, 2014

    2014

  38. [46]

    Lorenz and S

    R. Lorenz and S. Tull. Causal models in string diagrams , 2023. arXiv:2304.07638

    2023 arXiv

  39. [47]

    Mac Lane

    S. Mac Lane . C ategories for the W orking M athematician . Springer New York, NY, 1998

    1998

  40. [48]

    Maron, H

    H. Maron, H. Ben-Hamu, N. Shamir, and Y. Lipman. Invariant and E quivariant G raph N etworks. In International Conference on Learning Representations, 2019 a

    2019

  41. [49]

    Maron, E

    H. Maron, E. Fetaya, N. Segol, and Y. Lipman. On the U niversality of I nvariant N etworks. In Proceedings of the 36th International Conference on Machine Learning, volume 97 of Proceedings of Machine Learning Research, pages 4363--4371. PMLR, 09--15 Jun 2019 b

    2019

  42. [50]

    Maron, O

    H. Maron, O. Litany, G. Chechik, and E. Fetaya. On L earning S ets of S ymmetric E lements. In Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 6734--6744. PMLR, 13--18 Jul 2020

    2020

  43. [51]

    P. P. Martin. Representations of G raph T emperley-- L ieb A lgebras. Publications of the Research Institute for Mathematical Sciences, 26 0 (3): 0 485--503, 1990

    1990

  44. [52]

    P. P. Martin. T emperley-- L ieb A lgebras for N on-- P lanar S tatistical M echanics -- the P artition A lgebra C onstruction. Journal of Knot Theory and Its Ramifications, 03 0 (01): 0 51--82, 1994

    1994

  45. [53]

    P. P. Martin. The S tructure of the P artition A lgebras. Journal of Algebra, 183: 0 319--358, 1996

    1996

  46. [54]

    Pan and R

    H. Pan and R. Kondor. Permutation E quivariant L ayers for H igher O rder I nteractions. In Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, volume 151 of Proceedings of Machine Learning Research, pages 5987--6001. PMLR, 28--30 Mar 2022

    2022

  47. [55]

    Pearce-Crump

    E. Pearce-Crump. B rauer's G roup E quivariant N eural N etworks. In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, pages 27461--27482. PMLR, 23--29 Jul 2023 a

    2023

  48. [56]

    Pearce-Crump

    E. Pearce-Crump. How Jellyfish Characterise Alternating Group Equivariant Neural Networks . In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, pages 27483--27495. PMLR, 23--29 Jul 2023 b

    2023

  49. [57]

    Pearce-Crump

    E. Pearce-Crump. Connecting Permutation Equivariant Neural Networks and Partition Diagrams . In Proceedings of the 27th European Conference on Artificial Intelligence (ECAI 2024), volume 392 of Frontiers in Artificial Intelligence and Applications, pages 1511--1518. IOS Press, 2024

    2024

  50. [58]

    Pearce-Crump and W

    E. Pearce-Crump and W. Knottenbelt. Graph Automorphism Group Equivariant Neural Networks . In Proceedings of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine Learning Research, pages 40051--40077. PMLR, 21--27 Jul 2024

    2024

  51. [59]

    Petrache and S

    M. Petrache and S. Trivedi. Position Paper: Generalized Grammar Rules and Structure-Based Generalization Beyond Classical Equivariance for Lexical Tasks and Transduction , 2024. arXiv:2402.01629

    2024 arXiv

  52. [60]

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation . In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 652--660, 2017

    2017

  53. [61]

    Rahme, S

    J. Rahme, S. Jelassi, J. Bruna, and S. M. Weinberg. A Permutation-Equivariant Neural Network Architecture For Auction Design . Proceedings of the AAAI Conference on Artificial Intelligence, 35 0 (6): 0 5664--5672, 2021

    2021

  54. [62]

    Ravanbakhsh, J

    S. Ravanbakhsh, J. Schneider, and B. P \'o czos. Equivariance Through Parameter-Sharing . In Proceedings of the 34th International Conference on Machine Learning, volume 70, pages 2892--2901, 06--11 Aug 2017

    2017

  55. [63]

    E. Riehl. Category Theory in Context. Dover Publications, 2017

    2017

  56. [64]

    V. G. Satorras, E. Hoogeboom, and M. Welling. E(n) Equivariant Graph Neural Networks . In Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, pages 9323--9332. PMLR, 18--24 Jul 2021

    2021

  57. [65]

    A. Savage. S tring D iagrams and C ategorification , pages 3--36. Springer International Publishing, 2021

    2021

  58. [66]

    Thomas, T

    N. Thomas, T. Smidt, S. Kearnes, L. Yang, L. Li, K. Kohlhoff, and P. Riley. Tensor field networks: R otation-and translation-equivariant neural networks for 3d point clouds, 2018. arXiv:1802.08219

    2018 arXiv

  59. [67]

    S. Tull, R. Lorenz, S. Clark, I. Khan, and B. Coecke. Towards Compositional Interpretability for XAI , 2024. arXiv:2406.17583

    2024 arXiv

  60. [68]

    Turaev and A

    V. Turaev and A. Virelizier. Monoidal C ategories and T opological F ield T heory . Birkhäuser, 2017

    2017

  61. [69]

    van de Wetering

    J. van de Wetering. ZX-calculus for the working quantum computer scientist , 2020. arXiv:2012.13966

    2020 arXiv

  62. [70]

    Villar, D

    S. Villar, D. W. Hogg, K. Storey-Fisher, W. Yao, and B. Blum-Smith. Scalars are universal: E quivariant machine learning, structured like classical physics. In Advances in Neural Information Processing Systems, 2021

    2021

  63. [71]

    Weiler and G

    M. Weiler and G. Cesa. General E(2)-Equivariant Steerable CNNs . In Advances in Neural Information Processing Systems, volume 32, 2019

    2019

  64. [72]

    H. Wenzl. On the S tructure of B rauer's C entralizer A lgebras. Annals of Mathematics, 128 0 (1): 0 173--193, 1988

    1988

  65. [73]

    H. Weyl. The Classical Groups: Their Invariants and Representations. Princeton University Press, 1939

    1939

  66. [74]

    Zaheer, S

    M. Zaheer, S. Kottur, S. Ravanbakhsh, B. Poczos, R. R. Salakhutdinov, and A. J. Smola. Deep S ets. In Advances in Neural Information Processing Systems, volume 30, 2017

    2017

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.