Pith. sign in

REVIEW 3 major objections 5 minor 7 cited by

Neural Network-Augmented Pfaffian Wave-functions for Scalable Simulations of Interacting Fermions

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper introduces hidden fermion Pfaffian states (HFPS), a neural-network variational wave function for interacting fermions, and argues that HFPS achieves state-of-the-art variational accuracy across several phases of the…

desk verdict New restricted Pfaffian NQS with strong Hubbard benchmarks, but Eq. (14)'s WLOG decomposition is invalid and the 'generalizes HFDS' claim needs qualification. read the letter →

arxiv 2507.10705 v1 pith:USE7SIXI submitted 2025-07-14 cond-mat.str-el cond-mat.dis-nnquant-ph

Ao Chen , Zhou-Quan Wan , Anirvan Sengupta , Antoine Georges , Christopher Roth This is my paper · ORCID
classification cond-mat.str-elcond-mat.dis-nnquant-ph
keywords hiddenfermionPfaffianstateneuralquantumstatesHubbardmodelvariationalMonteCarlowavefunctionssuperconductivitystripeorderlow-rankupdates
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces hidden fermion Pfaffian states (HFPS), a neural-network variational wave function that combines a Pfaffian of visible and hidden fermion orbitals with a CNN-generated coupling and a Jastrow factor. The central claim is that this ansatz gives state-of-the-art variational accuracy for the two-dimensional Hubbard model in different phases, including a weakly correlated Fermi liquid, an attractive s-wave superconductor, a Mott insulator, and a stripe-ordered state, often reducing variational energy error by over an order of magnitude compared with earlier neural fermionic states. If true, HFPS is the first scalable neural quantum state that is both accurate enough to resolve tiny energy differences between competing phases and fast enough, with $O(N^3)$ per variational Monte Carlo step, to approach thermodynamically relevant system sizes. The paper also shows that HFPS captures both s-wave and d-wave pairing correlations, which matters for modeling unconventional superconductivity.

What carries the argument

The central object is the hidden fermion Pfaffian state, defined by a block Pfaffian matrix with visible-visible block $F^{vv}$, visible-hidden block $\tilde{F}^{vh}(n)$ generated by a convolutional neural network, and a constant hidden-hidden block $\tilde{F}^{hh}$, together with a neural Jastrow factor $J(n)$. Using the block Pfaffian identity, the paper rewrites HFPS as a Pfaffian neural backflow whose rank is controlled by the number of hidden fermions $\tilde{N}$, which enables low-rank updates that reduce the forward-pass cost from $O(N^3)$ to $O(N^2\tilde{N})$ and the full variational Monte Carlo step to $O(N^3)$. The constant $\tilde{F}^{hh}$ is justified by an anti-symmetric-matrix decomposition in Eq. (14), and the same construction supports sublattice and translation symmetrization.

What would settle it

On a small lattice, compare the optimal variational energy of HFPS with fixed $\tilde{F}_{hh}$ against the same architecture but with a freely varying configuration-dependent $F_{hh}(n)$; if the fixed version is strictly worse on any state whose hidden block changes singular values with $n$, then the without-loss-of-generality claim fails, while equal energies would confirm it.

Watch

Extended reading notes

Core claim

HFPS generalizes the hidden-fermion determinant state by replacing the determinant with a Pfaffian, so the wave function naturally contains Slater determinants, BCS paired states, and more general antisymmetric pairing states in one family. The paper's numerical experiments show that HFPS reaches relative energy errors near $10^{-5}$ on a $4\times4$ Hubbard lattice where exact diagonalization exists, and near $10^{-4}$ on an $8\times8$ half-filled lattice where auxiliary-field quantum Monte Carlo provides the reference, outperforming PP+RBM, HFDS, transformer neural backflow, and a simple fermionic projected entangled-pair state benchmark. In the $16\times4$ stripe-ordered Hubbard model at $1/8$ doping, HFPS obtains $E/N=-0.76413$, below the reported transformer neural backflow value, while reproducing spin-density-wave period 16 and charge-density-wave period 8 and showing short-range d-wave pairing. These results are presented as evidence that a scalable Pfaffian-based neural ansatz can resolve the small energy differences among competing fermionic phases and can serve as a practical tool for strongly correlated electrons.

Load-bearing premise

The derivation of Eq. (14) assumes that the hidden-hidden pairing block $F_{hh}(n)$ can be rotated to a constant anti-symmetric matrix "without loss of generality"; if that rotation actually restricts the configuration dependence of the wave function, the implemented HFPS is less expressive than the formal hidden-fermion Pfaffian family, and the numerical results would need to be reread as evidence about a narrower ansatz.

Editorial extensions

If this is right

  • HFPS reaches relative variational energy errors near $10^{-5}$ on small lattices with exact references and near $10^{-4}$ on $8\times8$ half-filled systems, placing it among the most accurate scalable neural fermionic methods reported.
  • The $O(N^3)$ per-step scaling, together with low-rank Pfaffian updates and sublattice symmetry, allows forward passes on lattices up to 400 sites, which the paper reports as a factor-of-$10^3$ speedup over the unaccelerated implementation.
  • In the stripe phase at $1/8$ doping, HFPS yields a lower variational energy than the recent transformer neural backflow result while simultaneously reproducing the expected spin and charge density wave periods and short-range d-wave pairing.
  • Because the Pfaffian family contains both Slater determinants and BCS states, HFPS can interpolate between metallic, superconducting, and stripe-ordered wave functions within a single architecture, making it a candidate tool for mapping phase diagrams of strongly correlated fermion models.

Reading between the lines

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

  • The paper does not test the t-$t'$ Hubbard model, but HFPS appears well suited to search for d-wave superconducting order there, since its Pfaffian structure explicitly accommodates pairing while the CNN provides the correlations that could stabilize or destroy long-range order.
  • The rank-controlled backflow interpretation suggests a tunable interpolation between cheap mean-field-like states and expensive full-backflow states: increasing $\tilde{N}$ should smoothly increase both expressivity and cost, a trade-off the paper does not systematically explore.
  • The constant-$\tilde{F}_{hh}$ assumption in Eq. (14) is directly testable: on a small lattice, one could compare the implemented HFPS against a version that lets $F_{hh}(n)$ vary freely; equal energies would confirm the without-loss-of-generality claim, while a gap would show that the implemented ansatz is narrower than the formal hidden-fermion Pfaffian family.
  • The demonstrated accuracy on the $16\times4$ stripe suggests that measuring generalized correlation functions in HFPS could provide variational estimates of effective low-energy quasiparticle interactions, a direction the paper mentions but does not pursue numerically.
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

3 major / 5 minor

Summary. The manuscript introduces the hidden fermion Pfaffian state (HFPS), a variational fermionic wave function that augments a Pfaffian mean-field ansatz with hidden-fermion blocks and Jastrow factors generated by convolutional neural networks. The authors derive Eq. (14) by a spectral decomposition of the configuration-dependent hidden-hidden pairing block, describe low-rank-update and sublattice-symmetry accelerations that reduce the forward-pass cost to O(N^2), and benchmark HFPS on the 2D Hubbard model in the weakly correlated regime (4x4, n=5/8), the attractive regime (8x8, n=7/8), the half-filled Mott insulator (8x8), and the stripe phase (L x 4, n=7/8, U=8). Reported variational energy errors reach 1e-4 to 1e-5 relative to ED, DQMC, or AFQMC references where available, and the 16x4 stripe energy is lower than a recent Transformer-NNBF result. The central claim is that HFPS generalizes HFDS and provides state-of-the-art variational accuracy in all studied regimes.

Significance. If the central claims hold, HFPS would be the most accurate scalable neural variational method for the 2D Hubbard model in the tested regimes, with energy errors close to exact references and an architecture that can be extended to large systems. The numerical benchmark program is strong: ED, DQMC, and AFQMC references are used where available; raw data are tabulated in Appendix G; and the complexity reductions are quantified in Fig. 3 and Table I. The theoretical claim that HFPS generalizes HFDS, however, relies on an unjustified spectral decomposition in Eq. (14), and the manuscript should be revised to specify precisely which variational family is implemented and to what extent the generalization claim is exact.

major comments (3)
  1. [Section II.B, Eq. (14)] The assertion that F_hh(n) = V^T(n) F̃_hh V(n) with constant F̃_hh and unitary V(n) holds 'without loss of generality' is not correct. For a real antisymmetric matrix A, unitary congruence A -> V^T A V preserves the singular values of A; hence a fixed F̃_hh can represent only those F_hh(n) whose singular values are independent of n. The HFDS form F_hh(n)=u_h(n) J u_h(n)^T in Eq. (13) generally has n-dependent singular values, so Eq. (14) is not an exact reformulation of Eq. (11) and the statement that HFPS generalizes HFDS is not established. The numerical results may still stand for the implemented ansatz, but the theoretical framing must be corrected, for instance by presenting Eq. (14) directly as the definition of the variational family actually used and characterizing its relation to HFDS as a restricted subclass rather than an exact equivalence.
  2. [Section II.B, Eq. (14) and Section II.C] There is an internal inconsistency in the treatment of J(n). The derivation defines J(n)=det(U(n))=det(V(n)), which for a real orthogonal V(n) takes values ±1 and for a complex unitary V(n) is a phase, whereas Section II.C states that J(n) is computed by the ANN as a generalized Jastrow factor with dedicated output channels and no stated unit-modulus constraint. If J(n) is an independent network output, Eq. (14) is not a consequence of Eq. (11) but a different, larger ansatz; if J(n) is constrained to equal det(V(n)), the ANN implementation overparameterizes it. The authors should clarify which construction is actually optimized and adjust the derivation and the claims about the relation between Eq. (11) and Eq. (14) accordingly.
  3. [Section III.C, Fig. 6] The comparison with fPEPS is weakened by the acknowledgment that the D=16 fPEPS result is obtained with simple-update optimization and 'not pushed to its limiting expressive power.' This is a fair caveat, but it means the 'surpassing other methods by more than an order of magnitude' statement in Section I relies on comparisons to non-optimized reference data for at least one method. The central HFPS-vs-AFQMC accuracy claim is not affected, but the manuscript should be more careful in distinguishing benchmarks that are state of the art for the respective methods from those that are not.
minor comments (5)
  1. [Section III.D] There is a typo, 'competetive', which should read 'competitive'.
  2. [Section II.D] The word 'utilizied' appears; it should read 'utilized'.
  3. [Fig. 1 caption] The caption contains the fragment '0 11 0' with no explanation; this appears to be a rendering artifact of the Fock-state labels and should be cleaned up.
  4. [Fig. 6] The figure caption notes that HFDS uses PBC/APBC while other methods use PBC on both sides; this boundary-condition difference should also be stated in the main text at the point where the comparison is discussed, not only in the caption.
  5. [Section III.C] The statement that a 4x4 sublattice unit cell is 'large enough to encode the translational symmetry breaking of the best mean-field state' is presented as an explanation of the observed similarity between the no-sublattice and sublattice-symmetrized results; it would be helpful to add a sentence noting that this is an empirical observation for the studied cases rather than a general guarantee.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: variational energies are minimized against the Hubbard Hamiltonian only, all accuracy benchmarks (ED, DQMC, AFQMC) are external, and the HFPS-generalizes-HFDS claim is proven by explicit parameter assignment.

full rationale

The derivation chain is self-contained. The ansatz is defined in Eqs. (8)-(11), and the rewrite to Eq. (14) uses the Pfaffian congruence identity (B5) with a spectral decomposition of the antisymmetric hidden block; the one contested point in this step — that Fhh(n) = V^T(n) F̃hh V(n) with constant F̃hh holds 'without loss of generality' — is a validity question about expressivity (unitary congruence preserves the singular values of an antisymmetric matrix), not a circular reduction, because the variational energies reported are obtained by minimizing ⟨H⟩ over the implemented family with no reference to any benchmark value. The central generalization claim (HFPS ⊃ HFDS, Section II.B) is proven by explicit assignment Fvv = uvJ(uv)^T, Fvh(n) = uvJuh(n)^T, Fhh(n) = uh(n)Juh(n)^T, plus the full-rank-Fvv counterexample for the converse; no fitted quantity enters. All accuracy benchmarks are external and independent: exact diagonalization (QuSpin) for the 4×4 case, DQMC for the 8×8 attractive case, and AFQMC for the 8×8 half-filled case, and the paper selects boundary conditions per method so the comparison is not distorted. Hyperparameters (mean-field initialization at |U|=3, d-wave field Δ=0.2 on 16×4) are chosen by variational-energy improvement within the paper's own objective — the paper explicitly reports that the pairing-free initialization yields the higher energy E/N = −0.76256 — so they are not fitted to match reference energies. Self-citations ([41], [42], [90], [93], [95]) supply optimizers, architectures, and software that do not carry the central accuracy claim; the HFDS comparison [53] shares an author, but the improvement over HFDS is corroborated by the independent ED energies in Fig. 4 and by AFQMC/DQMC benchmarks elsewhere. The paper's own caveats (fPEPS 'not pushed to its limiting expressive power'; sublattice HFPS having 'lower expressive power'; Appendix B Pfaffian properties listed 'without proof') are honesty about standard settings, not circular support. Verdict: no circular step; the flagged Eq. (14) 'WLOG' issue is a correctness/framing concern and should be weighed there, not as circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method's central claim depends on standard Pfaffian mathematics, plus several hand-chosen hyperparameters and two structural assumptions. The most fragile is the constant F̃hh decomposition, which restricts the ansatz but may not affect the numerical benchmarks.

free parameters (5)
  • d-wave pairing field Δ in mean-field initialization = 0.2
    Chosen by hand for the 16x4 stripe simulation (Section III.D) to improve convergence; not derived from first principles.
  • Mean-field initialization interaction U_init = |U|=3
    Used to pre-train the Slater/BCS/Thouless state for all target U (Section II.C), based on Ref [70] overlap argument; a hand-chosen hyperparameter.
  • Number of hidden fermions Ñ = 14
    28 of 32 CNN channels form F̃vh with Ñ=14 (Section II.C); hand-chosen balance of expressivity and cost.
  • Sublattice unit cell size = 4x4
    Chosen for the Mott and stripe simulations based on mean-field symmetry (Sections II.C and III.C); affects the symmetry projection cost.
  • CNN hyperparameters = 16 layers, 8 residual blocks, 3x3 kernels, 32 channels, ~150k params
    Adopted from Ref [42,69]; not tuned on the target problems, so they are inherited rather than fitted, but they influence results.
assumptions (4)
  • standard math Pfaffian algebraic identities (B4-B10), including block Pfaffian formula and Pfaffian ratio under low-rank update
    Used throughout appendices D and E; standard results stated without proof.
  • ad hoc to paper Spectral decomposition of real anti-symmetric matrices into V^T F̃hh V with constant F̃hh and all n-dependence in V
    Invoked in Section II.B around Eq (14) with 'without loss of generality'; not generally valid because singular values are invariant under orthogonal congruence, so this restricts Fhh(n).
  • domain assumption Mean-field states trained at |U|=3 provide good initialization for the full interacting problem at various U
    Used in Section II.C; based on empirical observation from Ref [70] and the authors' own tests; if the overlap is poor, SR may converge to local minima.
  • ad hoc to paper The 4x4 sublattice unit cell is large enough to encode the relevant symmetry breaking of the best mean-field state
    Stated in Section III.C; chosen from mean-field density matrix symmetry, not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Network-Augmented Pfaffian Wave-functions for Scalable Simulations of Interacting Fermions." pith.science (2026). https://pith.science/paper/USE7SIXI

@misc{pith2026250710705,
  author       = {Pith},
  title        = {Pith review of: Neural Network-Augmented Pfaffian Wave-functions for Scalable Simulations of Interacting Fermions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/USE7SIXI}},
  note         = {Machine review of arXiv:2507.10705}
}
read the original abstract

Developing accurate numerical methods for strongly interacting fermions is crucial for improving our understanding of various quantum many-body phenomena, especially unconventional superconductivity. Recently, neural quantum states have emerged as a promising approach for studying correlated fermions, highlighted by the hidden fermion and backflow methods, which use neural networks to model corrections to fermionic quasiparticle orbitals. In this work, we expand these ideas to the space of Pfaffians, a wave-function that naturally expresses superconducting pairings, and propose the hidden fermion Pfaffian state (HFPS), which flexibly represents both unpaired and superconducting phases and scales to large systems with favorable asymptotic complexity. In our numerical experiments, HFPS provides state-of-the-art variational accuracy in different regimes of both the attractive and repulsive Hubbard models. We show that the HFPS is able to capture both s-wave and d-wave pairing, and therefore may be a useful tool for modeling phases with unconventional superconductivity.

Figures

Figures reproduced from arXiv: 2507.10705 by the authors.

Figure 1
Figure 1. FIG. 1. Illustration of HFPS architecture in Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Illustration of HFPS architecture with sublattice and translation symmetry as an extension of the architecture in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Comparison of the time cost of a HFPS forward pass [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: FIG. 5. The pair-pair correlation [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Relative error of energy [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Variational energy in the spin- and charge-ordered [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. The spin, charge, and pair correlations in the 16 [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Large scale neural quantum states reveal the interplay between superconductivity and quantum criticality in the Hofstadter-Hubbard model

    cond-mat.str-el 2026-08 conditional novelty 7.0 of 10

    Neural quantum state simulations of the Hofstadter-Hubbard model show a continuous IQH to chiral spin liquid transition and a topological superconductor whose stiffness is enhanced near the quantum critical point.

  2. Exact Neural-Network Representations of the Motzkin States

    cond-mat.str-el 2026-07 conditional novelty 7.0 of 10

    Exact training-free neural-network representations for colorless and colorful Motzkin states are constructed for RNN, FNN, CNN, and transformer architectures.

  3. Engineering two-body interaction for the Moore-Read State

    cond-mat.str-el 2026-07 conditional novelty 6.0 of 10

    Gradient-optimized two-body Haldane pseudopotentials yield Moore–Read Pfaffian overlaps above 99% up to 12 electrons and collapse to a nearly universal exponential short-range profile.

  4. Attention-Based Foundation Model for Quantum States

    cond-mat.str-el 2025-12 conditional novelty 6.0 of 10

    A single attention-based transformer, conditioned on the parameters (V/t,N), learns and predicts ground-state wavefunctions of the t-V model across a 2D parameter plane from a few exact-diagonalization training points.

  5. Comparing Symmetrized Determinant Neural Quantum States for the Hubbard Model

    cond-mat.str-el 2025-10 conditional novelty 6.0 of 10

    For the doped square-lattice Hubbard model, hidden-fermion and backflow neural quantum states with a Vision Transformer backbone reach nearly equal variational energies; translation-equivariant attention is outperform...

  6. Group Convolutional Neural Network for the Low-Energy Spectrum in the Quantum Dimer Model

    cond-mat.dis-nn 2025-05 conditional novelty 6.0 of 10

    Irrep-resolved GCNN variational energies indicate a 4-fold degenerate columnar ground state for V <= 0.4 in the square-lattice quantum dimer model, shifting possible plaquette or mixed ordering to 0.4 < V < 1.

  7. Optimization dynamics of Transformer backflow neural quantum states for the two-dimensional Hubbard model

    cond-mat.str-el 2026-07 conditional novelty 4.0 of 10

    For Transformer backflow neural quantum states on the 4x4 doped Hubbard model, larger width, more determinant channels, and larger Monte Carlo batches improve final accuracy, and a less restrictive update-norm thresho...

Reference graph

Works this paper leans on

106 extracted references · 51 canonical work pages · cited by 7 Pith papers

  1. [1]

    The single-particle eigenstates of this Hamiltonian define a quasi-particle transformation ˆγ† α = MX p upαˆc† p

    Determinant state Consider a bilinear Hamiltonian with a conserved num- ber of electrons ˆH0 = MX p,q tpqˆc† pˆcq, (C1) where the indices p and q run over all M orbitals of the system, including sites and spins. The single-particle eigenstates of this Hamiltonian define a quasi-particle transformation ˆγ† α = MX p upαˆc† p. (C2) The Slater determinant sta...

  2. [2]

    The new quasiparticles are then defined as ˆγ† α = X p uv pαˆc† p + X ˜p uh ˜pα ˆd† ˜p (C5) where uv and uh denote the transformation matrices, and ˆd† ˜p denote hidden particles

    Adding hidden fermions The HFDS generalizes the Slater determinant to an enlarged Hilbert space with M + ˜M orbitals and N + ˜N fermions. The new quasiparticles are then defined as ˆγ† α = X p uv pαˆc† p + X ˜p uh ˜pα ˆd† ˜p (C5) where uv and uh denote the transformation matrices, and ˆd† ˜p denote hidden particles. Then the wave-function component of HFD...

  3. [3]

    0 −u1 0

    Rank-1 update Consider two Fock statesn and n0 only different by one fermion hopping, and M = n⋆F⋆n and M0 = n0 ⋆F⋆n0 14 only different at the j’th row and j’th column, then M − M0 =   0 ... 0 −u1 0 ... 0 ... ... ... ... ... 0 ... 0 −uj−1 0 ... 0 u1 ... uj−1 0 uj+1 ... uNe 0 ... 0 −uj+1 0 ... 0 ... ... ... ... ... 0 ... 0 −uNe 0 ... 0  ...

  4. [4]

    Then we still have M − M0 = −vJkvT , (E6) and R = Jk + vT M−1 0 v, (E7) pfM pfM0 = pfR pfJk , (E8) M−1 = M−1 0 + (M−1 0 v)R−1(M−1 0 v)T , (E9) where R is a 2k ×2k matrix

    Rank- k update When M and M0 are different at j1, j2, ..., jk rows and columns, one can make v an N × 2k matrix with the first k columns given by the non-zero rows of M − M0, and the last k columns given by ei,n = δi,jn . Then we still have M − M0 = −vJkvT , (E6) and R = Jk + vT M−1 0 v, (E7) pfM pfM0 = pfR pfJk , (E8) M−1 = M−1 0 + (M−1 0 v)R−1(M−1 0 v)T...

  5. [5]

    (14) and the block Pfaffian formula in Eq

    LRU for HFPS Combining the HFPS wave-function in Eq. (14) and the block Pfaffian formula in Eq. (B9), one obtains ψ(n) = J(n) × pf(n ⋆ Fvv ⋆ n) × pf[ ˜Fhh + (˜Fvh(n)T ⋆ n)(n ⋆ Fvv ⋆ n)−1(n ⋆ ˜Fvh(n))]. (E10) J(n) cannot be computed by local updates, but it does not involve the Pfaffian complexity. The first Pfaffian pf(n ⋆ Fvv ⋆ n) can be computed by LRU ...

  6. [6]

    We consider a group of symme- try operations g ∈ G over which a fermion Hamiltonian is invariant

    Symmetry in fermion systems For simplicity, we focus on spinless fermions in this subsection, while the discussion can be directly general- ized to spinful fermions. We consider a group of symme- try operations g ∈ G over which a fermion Hamiltonian is invariant. This group can include translations, as well 15 as point group operations, which include rota...

  7. [7]

    As shown in Fig

    T ranslation symmetry in HFPS Consider the translated HFPS wave-function ψ(Tyn) = J(Tyn) ×pf (Tyn) ⋆ Fvv ⋆ (Tyn) ( Tyn) ⋆ ˜Fvh(Tyn) − ˜Fvh(Tyn)T ⋆ (Tyn) ˜Fhh , (F6) where we have three y dependent parts, namely J(Tyn), (Tyn) ⋆ Fvv ⋆ (Tyn), and ( Tyn) ⋆ ˜Fvh(Tyn). As shown in Fig. 2, the CNN output for the Jastrow factor is an array J , and for each transl...

  8. [8]

    We instead allow symmetry breaking within an enlarged unit cell, which is informed by our prior knowledge of the problem

    Sublattice Symmetry In practice, we rarely choose F vv to be fully transla- tionally invariant, as this inhibits our ability to model symmetry-broken phases. We instead allow symmetry breaking within an enlarged unit cell, which is informed by our prior knowledge of the problem. Typically, we choose this unit cell by first training a Slater determinant An...

Show all 106 references
  1. [9]

    The GCNN generalizes the CNN from the translation group to general non-abelian groups G

    HFPS + GCNN Hidden fermion Pfaffian states can also be combined with group convolutional neural networks (GCNNs) [41, 96, 97] instead of CNNs. The GCNN generalizes the CNN from the translation group to general non-abelian groups G. The feature maps of a GCNN encode a represent...

  2. [10]

    J. G. Bednorz and K. A. M¨ uller, Possible hightc su- perconductivity in the ba-la-cu-o system, Zeitschrift f¨ ur Physik B Condensed Matter 64, 189 (1986)

    1986

  3. [11]

    M. K. Wu, J. R. Ashburn, C. J. Torng, P. H. Hor, R. L. Meng, L. Gao, Z. J. Huang, Y. Q. Wang, and C. W. Chu, Superconductivity at 93 k in a new mixed-phase y-ba-cu-o compound system at ambient pressure, Phys. Rev. Lett. 58, 908 (1987)

    1987

  4. [12]

    M. Qin, T. Sch¨ afer, S. Andergassen, P. Corboz, and E. Gull, The hubbard model: A computational perspec- tive, Annual Review of Condensed Matter Physics 13, 275 (2022)

    2022

  5. [13]

    D. P. Arovas, E. Berg, S. A. Kivelson, and S. Raghu, The hubbard model, Annual Review of Condensed Mat- ter Physics 13, 239 (2022)

    2022

  6. [14]

    Becca and S

    F. Becca and S. Sorella, Quantum Monte Carlo Ap- proaches for Correlated Systems (Cambridge University Press, 2017)

    2017

  7. [15]

    D. Wu, R. Rossi, F. Vicentini, N. Astrakhantsev, F. Becca, X. Cao, J. Carrasquilla, F. Ferrari, A. Georges, M. Hibat-Allah, M. Imada, A. M. L¨ auchli, G. Mazzola, A. Mezzacapo, A. Millis, J. R. Moreno, T. Neupert, Y. Nomura, J. Nys, O. Parcollet, R. Pohle, I. Romero, M. Schmid...

    2024

  8. [16]

    S. R. White, Density matrix formulation for quantum renormalization groups, Phys. Rev. Lett.69, 2863 (1992)

    1992

  9. [17]

    S. R. White, Density-matrix algorithms for quantum renormalization groups, Phys. Rev. B 48, 10345 (1993)

    1993

  10. [18]

    Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326, 96 (2011)

    U. Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326, 96 (2011)

    2011

  11. [19]

    S. R. White and D. J. Scalapino, Density matrix renor- malization group study of the striped phase in the 2d t − J model, Phys. Rev. Lett. 80, 1272 (1998)

    1998

  12. [20]

    Jiang and T

    H.-C. Jiang and T. P. Devereaux, Superconductivity in the doped hubbard model and its interplay with next- nearest hopping, Science 365, 1424 (2019)

    2019

  13. [21]

    Jiang and S

    H.-C. Jiang and S. A. Kivelson, Stripe order enhanced su- perconductivity in the hubbard model, Proceedings of the National Academy of Sciences 119, e2109406119 (2022)

    2022

  14. [22]

    Corboz and G

    P. Corboz and G. Vidal, Fermionic multiscale entangle- ment renormalization ansatz, Phys. Rev. B 80, 165129 (2009)

    2009

  15. [23]

    Corboz, R

    P. Corboz, R. Or´ us, B. Bauer, and G. Vidal, Simula- tion of strongly correlated fermions in two spatial di- mensions with fermionic projected entangled-pair states, Phys. Rev. B 81, 165104 (2010)

    2010

  16. [24]

    C. V. Kraus, N. Schuch, F. Verstraete, and J. I. Cirac, Fermionic projected entangled pair states, Phys. Rev. A 81, 052338 (2010)

    2010

  17. [25]

    Mortier, L

    Q. Mortier, L. Devos, L. Burgelman, B. Vanhecke, N. Bultinck, F. Verstraete, J. Haegeman, and L. Van- derstraeten, Fermionic tensor network methods, SciPost Phys. 18, 012 (2025)

    2025

  18. [26]

    W.-Y. Liu, H. Zhai, R. Peng, Z.-C. Gu, and G. K.-L. Chan, Accurate simulation of the hubbard model with fi- nite fermionic projected entangled pair states, Phys. Rev. Lett. 134, 256502 (2025)

    2025

  19. [27]

    Blankenbecler, D

    R. Blankenbecler, D. J. Scalapino, and R. L. Sugar, Monte carlo calculations of coupled boson-fermion sys- tems. i, Phys. Rev. D 24, 2278 (1981)

    1981

  20. [28]

    J. E. Hirsch, Two-dimensional hubbard model: Numeri- cal simulation study, Phys. Rev. B 31, 4403 (1985)

    1985

  21. [29]

    S. R. White, D. J. Scalapino, R. L. Sugar, E. Y. Loh, J. E. Gubernatis, and R. T. Scalettar, Numerical study of the two-dimensional hubbard model, Phys. Rev. B 40, 506 (1989)

    1989

  22. [30]

    D. J. Scalapino, S. R. White, and S. Zhang, Insulator, metal, or superconductor: The criteria, Phys. Rev. B 47, 7995 (1993)

    1993

  23. [31]

    Assaad and H

    F. Assaad and H. Evertz, World-line and determinan- tal quantum monte carlo methods for spins, phonons and electrons, in Computational Many-Particle Physics , edited by H. Fehske, R. Schneider, and A. Weiße (Springer Berlin Heidelberg, Berlin, Heidelberg, 2008) pp. 277–356

    2008

  24. [32]

    Sugiyama and S

    G. Sugiyama and S. Koonin, Auxiliary field monte-carlo for quantum many-body ground states, Annals of Physics 168, 1 (1986)

    1986

  25. [33]

    Zhang, J

    S. Zhang, J. Carlson, and J. E. Gubernatis, Constrained path monte carlo method for fermion ground states, Phys. Rev. B 55, 7464 (1997)

    1997

  26. [34]

    Shi and S

    H. Shi and S. Zhang, Symmetry in auxiliary-field quan- tum monte carlo calculations, Phys. Rev. B 88, 125132 (2013)

    2013

  27. [35]

    Troyer and U.-J

    M. Troyer and U.-J. Wiese, Computational complexity and fundamental limitations to fermionic quantum monte carlo simulations, Phys. Rev. Lett. 94, 170201 (2005)

    2005

  28. [36]

    Ceperley, G

    D. Ceperley, G. V. Chester, and M. H. Kalos, Monte carlo simulation of a many-fermion study, Phys. Rev. B 16, 3081 (1977)

    1977

  29. [37]

    Yokoyama and H

    H. Yokoyama and H. Shiba, Variational monte-carlo studies of hubbard model. i, Journal of the Physical So- ciety of Japan 56, 1490 (1987)

    1987

  30. [38]

    Yokoyama and H

    H. Yokoyama and H. Shiba, Variational monte-carlo studies of hubbard model. ii, Journal of the Physical So- ciety of Japan 56, 3582 (1987)

    1987

  31. [39]

    Yokoyama and H

    H. Yokoyama and H. Shiba, Variational monte-carlo studies of hubbard model. iii. intersite correlation effects, Journal of the Physical Society of Japan 59, 3669 (1990)

    1990

  32. [40]

    Gros, Superconductivity in correlated wave functions, Phys

    C. Gros, Superconductivity in correlated wave functions, Phys. Rev. B 38, 931 (1988)

    1988

  33. [41]

    Gros, Physics of projected wavefunctions, Annals of Physics 189, 53 (1989)

    C. Gros, Physics of projected wavefunctions, Annals of Physics 189, 53 (1989)

    1989

  34. [42]

    Paramekanti, M

    A. Paramekanti, M. Randeria, and N. Trivedi, Projected wave functions and high temperature superconductivity, Phys. Rev. Lett. 87, 217002 (2001)

    2001

  35. [43]

    Sorella, G

    S. Sorella, G. B. Martins, F. Becca, C. Gazza, L. Capri- otti, A. Parola, and E. Dagotto, Superconductivity in the two-dimensional t −J model, Phys. Rev. Lett. 88, 117002 (2002)

    2002

  36. [44]

    Paramekanti, M

    A. Paramekanti, M. Randeria, and N. Trivedi, High- Tc superconductors: A variational theory of the supercon- ducting state, Phys. Rev. B 70, 054504 (2004)

    2004

  37. [45]

    L. F. Tocchio, F. Becca, A. Parola, and S. Sorella, Role of backflow correlations for the nonmagnetic phase of the t–t ′ hubbard model, Phys. Rev. B 78, 041101 (2008)

    2008

  38. [46]

    Tahara and M

    D. Tahara and M. Imada, Variational monte carlo method combined with quantum-number projection and 18 multi-variable optimization, Journal of the Physical So- ciety of Japan 77, 114701 (2008)

    2008

  39. [47]

    Misawa, S

    T. Misawa, S. Morita, K. Yoshimi, M. Kawamura, Y. Mo- toyama, K. Ido, T. Ohgoe, M. Imada, and T. Kato, mvmc—open-source software for many-variable varia- tional monte carlo method, Computer Physics Commu- nications 235, 447 (2019)

    2019

  40. [48]

    Carleo and M

    G. Carleo and M. Troyer, Solving the quantum many- body problem with artificial neural networks, Science 355, 602 (2017)

    2017

  41. [49]

    Nomura and M

    Y. Nomura and M. Imada, Dirac-type nodal spin liq- uid revealed by refined quantum many-body solver using neural-network wave function, correlation ratio, and level spectroscopy, Phys. Rev. X 11, 031034 (2021)

    2021

  42. [50]

    C. Roth, A. Szab´ o, and A. H. MacDonald, High-accuracy variational monte carlo for frustrated magnets with deep neural networks, Phys. Rev. B 108, 054410 (2023)

    2023

  43. [51]

    Chen and M

    A. Chen and M. Heyl, Empowering deep neural quantum states through efficient optimization, Nature Physics 20, 1476 (2024)

    2024

  44. [52]

    L. L. Viteritti, R. Rende, A. Parola, S. Goldt, and F. Becca, Transformer wave function for two dimensional frustrated magnets: Emergence of a spin-liquid phase in the shastry-sutherland model, Phys. Rev. B 111, 134411 (2025)

    2025

  45. [53]

    K. Choo, A. Mezzacapo, and G. Carleo, Fermionic neural-network states for ab-initio electronic structure, Nature Communications 11, 2368 (2020)

    2020

  46. [54]

    D. Pfau, J. S. Spencer, A. G. D. G. Matthews, and W. M. C. Foulkes, Ab initio solution of the many-electron schr¨ odinger equation with deep neural networks, Phys. Rev. Res. 2, 033429 (2020)

    2020

  47. [55]

    Hermann, Z

    J. Hermann, Z. Sch¨ atzle, and F. No´ e, Deep-neural- network solution of the electronic schr¨ odinger equation, Nature Chemistry 12, 891 (2020)

    2020

  48. [56]

    von Glehn, J

    I. von Glehn, J. S. Spencer, and D. Pfau, A self- attention ansatz for ab-initio quantum chemistry (2023), arXiv:2211.13672 [physics.chem-ph]

    2023 arXiv

  49. [57]

    D. Pfau, S. Axelrod, H. Sutterud, I. von Glehn, and J. S. Spencer, Accurate computation of quantum ex- cited states with neural networks, Science 385, eadn0137 (2024)

    2024

  50. [58]

    Pescia, J

    G. Pescia, J. Nys, J. Kim, A. Lovato, and G. Carleo, Message-passing neural quantum states for the homoge- neous electron gas, Phys. Rev. B 110, 035108 (2024)

    2024

  51. [59]

    Smith, Y

    C. Smith, Y. Chen, R. Levy, Y. Yang, M. A. Morales, and S. Zhang, Unified variational approach description of ground-state phases of the two-dimensional electron gas, Phys. Rev. Lett. 133, 266504 (2024)

    2024

  52. [60]

    Nomura, A

    Y. Nomura, A. S. Darmawan, Y. Yamaji, and M. Imada, Restricted boltzmann machine learning for solving strongly correlated quantum systems, Phys. Rev. B 96, 205152 (2017)

    2017

  53. [61]

    Luo and B

    D. Luo and B. K. Clark, Backflow transformations via neural networks for quantum many-body wave functions, Phys. Rev. Lett. 122, 226401 (2019)

    2019

  54. [62]

    J. R. Moreno, G. Carleo, A. Georges, and J. Stokes, Fermionic wave functions from neural-network con- strained hidden states, Proceedings of the National Academy of Sciences 119, e2122059119 (2022)

    2022

  55. [63]

    Bardeen, L

    J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Micro- scopic theory of superconductivity, Phys. Rev. 106, 162 (1957)

    1957

  56. [64]

    Thouless, Vibrational states of nuclei in the random phase approximation, Nuclear Physics 22, 78 (1961)

    D. Thouless, Vibrational states of nuclei in the random phase approximation, Nuclear Physics 22, 78 (1961)

    1961

  57. [65]

    A. J. Coleman, Structure of fermion density matrices. ii. antisymmetrized geminal powers, Journal of Mathemat- ical Physics 6, 1425 (1965)

    1965

  58. [66]

    Bouchaud, J.P., Georges, A., and Lhuillier, C., Pair wave functions for strongly correlated fermions and their deter- minantal representation, J. Phys. France 49, 553 (1988)

    1988

  59. [67]

    P. R. Surj´ an, An introduction to the theory of geminals, in Correlation and Localization , edited by P. R. Surj´ an, R. J. Bartlett, F. Bog´ ar, D. L. Cooper, B. Kirtman, W. Klopper, W. Kutzelnigg, N. H. March, P. G. Mezey, H. M¨ uller, J. Noga, J. Paldus, J. Pipek, M. Raimo...

    1999

  60. [68]

    Moore and N

    G. Moore and N. Read, Nonabelions in the fractional quantum hall effect, Nuclear Physics B 360, 362 (1991)

    1991

  61. [69]

    Storni, R

    M. Storni, R. H. Morf, and S. Das Sarma, Fractional quantum hall state at ν = 5 2 and the moore-read pfaffian, Phys. Rev. Lett. 104, 076803 (2010)

    2010

  62. [70]

    R. G. Xu, T. Okubo, S. Todo, and M. Imada, Optimized implementation for calculation and fast-update of pfaf- fians installed to the open-source fermionic variational solver mvmc, Computer Physics Communications 277, 108375 (2022)

    2022

  63. [71]

    B. Fore, J. M. Kim, G. Carleo, M. Hjorth-Jensen, A. Lovato, and M. Piarulli, Dilute neutron star matter from neural-network quantum states, Phys. Rev. Res. 5, 033062 (2023)

    2023

  64. [72]

    J. Kim, G. Pescia, B. Fore, J. Nys, G. Carleo, S. Gan- dolfi, M. Hjorth-Jensen, and A. Lovato, Neural-network quantum states for ultra-cold fermi gases, Communica- tions Physics 7, 148 (2024)

    2024

  65. [73]

    Gao and S

    N. Gao and S. G¨ unnemann, Neural pfaffians: Solving many many-electron schr¨ odinger equations, inAdvances in Neural Information Processing Systems , Vol. 37, edited by A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang (Curran Asso- ciates, Inc...

    2024

  66. [74]

    B. Fore, J. Kim, M. Hjorth-Jensen, and A. Lovato, Inves- tigating the crust of neutron stars with neural-network quantum states, Communications Physics 8, 108 (2025)

    2025

  67. [75]

    K. Choo, T. Neupert, and G. Carleo, Two-dimensional frustrated J1−J2 model studied with neural network quantum states, Phys. Rev. B 100, 125124 (2019)

    2019

  68. [76]

    Liang, M

    X. Liang, M. Li, Q. Xiao, J. Chen, C. Yang, H. An, and L. He, Deep learning representations for quantum many-body systems on heterogeneous hardware, Machine Learning: Science and Technology 4, 015035 (2023)

    2023

  69. [77]

    L. L. Viteritti, R. Rende, and F. Becca, Transformer vari- ational wave functions for frustrated quantum spin sys- tems, Phys. Rev. Lett. 130, 236401 (2023)

    2023

  70. [78]

    A. Chen, V. D. Naik, and M. Heyl, Convolutional trans- former wave functions (2025), arXiv:2503.10462 [cond- mat.dis-nn]

    2025 arXiv

  71. [79]

    Zheng, C.-M

    B.-X. Zheng, C.-M. Chung, P. Corboz, G. Ehlers, M.-P. Qin, R. M. Noack, H. Shi, S. R. White, S. Zhang, and G. K.-L. Chan, Stripe order in the underdoped region of the two-dimensional hubbard model, Science 358, 1155 (2017)

    2017

  72. [80]

    Sorella, Green function monte carlo with stochastic reconfiguration, Phys

    S. Sorella, Green function monte carlo with stochastic reconfiguration, Phys. Rev. Lett. 80, 4558 (1998). 19

    1998

  73. [81]

    Sorella, M

    S. Sorella, M. Casula, and D. Rocca, Weak binding be- tween two aromatic rings: Feeling the van der waals at- traction by quantum monte carlo methods, The Journal of Chemical Physics 127, 014105 (2007)

    2007

  74. [82]

    Mazzola, A

    G. Mazzola, A. Zen, and S. Sorella, Finite-temperature electronic simulations without the Born-Oppenheimer constraint, The Journal of Chemical Physics 137, 10.1063/1.4755992 (2012)

    2012 doi

  75. [83]

    Goldshlager, N

    G. Goldshlager, N. Abrahamsen, and L. Lin, A kaczmarz- inspired approach to accelerate the optimization of neu- ral network wavefunctions, Journal of Computational Physics 516, 113351 (2024)

    2024

  76. [84]

    Romero, J

    I. Romero, J. Nys, and G. Carleo, Spectroscopy of two- dimensional interacting lattice electrons using symmetry- aware neural backflow transformations, Communications Physics 8, 46 (2025)

    2025

  77. [85]

    Q. Chen, J. Stajic, S. Tan, and K. Levin, Bcs–bec crossover: From high temperature superconductors to ul- tracold superfluids, Physics Reports 412, 1 (2005)

    2005

  78. [86]

    Y. Gu, W. Li, H. Lin, B. Zhan, R. Li, Y. Huang, D. He, Y. Wu, T. Xiang, M. Qin, L. Wang, and D. Lv, Solving the hubbard model with neural quantum states (2025), arXiv:2507.02644 [cond-mat.str-el]

    2025 arXiv

  79. [87]

    H. Shi, S. Chiesa, and S. Zhang, Ground-state proper- ties of strongly interacting fermi gases in two dimensions, Phys. Rev. A 92, 033603 (2015)

    2015

  80. [88]

    Liu, S.-J

    W.-Y. Liu, S.-J. Dong, Y.-J. Han, G.-C. Guo, and L. He, Gradient optimization of finite projected entangled pair states, Phys. Rev. B 95, 195154 (2017)

    2017

  81. [89]

    Liu, Y.-Z

    W.-Y. Liu, Y.-Z. Huang, S.-S. Gong, and Z.-C. Gu, Accu- rate simulation for finite projected entangled pair states in two dimensions, Phys. Rev. B 103, 235155 (2021)

    2021

  82. [90]

    Zhou, Z.-W

    Y.-T. Zhou, Z.-W. Zhou, and X. Liang, Solving fermi- hubbard-type models by tensor representations of back- flow corrections, Phys. Rev. B 109, 245107 (2024)

    2024

  83. [91]

    J. P. F. LeBlanc, A. E. Antipov, F. Becca, I. W. Bulik, G. K.-L. Chan, C.-M. Chung, Y. Deng, M. Ferrero, T. M. Henderson, C. A. Jim´ enez-Hoyos, E. Kozik, X.-W. Liu, A. J. Millis, N. V. Prokof’ev, M. Qin, G. E. Scuseria, H. Shi, B. V. Svistunov, L. F. Tocchio, I. S. Tupitsyn, ...

    2015

  84. [92]

    Qin, C.-M

    M. Qin, C.-M. Chung, H. Shi, E. Vitali, C. Hubig, U. Schollw¨ ock, S. R. White, and S. Zhang (Simons Col- laboration on the Many-Electron Problem), Absence of superconductivity in the pure two-dimensional hubbard model, Phys. Rev. X 10, 031016 (2020)

    2020

  85. [93]

    Xu, C.-M

    H. Xu, C.-M. Chung, M. Qin, U. Schollw¨ ock, S. R. White, and S. Zhang, Coexistence of superconductivity with par- tially filled stripes in the hubbard model, Science 384, eadh7691 (2024)

    2024

  86. [94]

    Wietek, Fragmented cooper pair condensation in striped superconductors, Phys

    A. Wietek, Fragmented cooper pair condensation in striped superconductors, Phys. Rev. Lett. 129, 177001 (2022)

    2022

  87. [95]

    Sorella, Systematically improvable mean-field varia- tional ansatz for strongly correlated systems: Applica- tion to the hubbard model, Phys

    S. Sorella, Systematically improvable mean-field varia- tional ansatz for strongly correlated systems: Applica- tion to the hubbard model, Phys. Rev. B 107, 115133 (2023)

    2023

  88. [96]

    Rende, L

    R. Rende, L. L. Viteritti, L. Bardone, F. Becca, and S. Goldt, A simple linear algebra identity to optimize large-scale neural network quantum states, Communica- tions Physics 7, 260 (2024)

    2024

  89. [97]

    Weinberg and M

    P. Weinberg and M. Bukov, QuSpin: a Python package for dynamics and exact diagonalisation of quantum many body systems part I: spin chains, SciPost Phys. 2, 003 (2017)

    2017

  90. [98]

    Weinberg and M

    P. Weinberg and M. Bukov, QuSpin: a Python package for dynamics and exact diagonalisation of quantum many body systems. Part II: bosons, fermions and higher spins, SciPost Phys. 7, 020 (2019)

    2019

  91. [99]

    Chen and C

    A. Chen and C. Roth, Quantax: Flexible neural quantum states based on QuSpin, JAX, and Equinox (2025)

    2025

  92. [100]

    Bradbury, R

    J. Bradbury, R. Frostig, P. Hawkins, M. J. Johnson, C. Leary, D. Maclaurin, G. Necula, A. Paszke, J. Van- derPlas, S. Wanderman-Milne, and Q. Zhang, JAX: com- posable transformations of Python+NumPy programs (2018)

    2018

  93. [101]

    Kidger and C

    P. Kidger and C. Garcia, Equinox: neural networks in jax via callable pytrees and filtered transformations (2021), arXiv:2111.00254 [cs.LG]

    2021 arXiv

  94. [102]

    Chen and C

    A. Chen and C. Roth, Fast low-rank update of matrix determinants and pfaffians in JAX (2025)

    2025

  95. [103]

    Liu and B

    Z. Liu and B. K. Clark, Unifying view of fermionic neural network quantum states: From neural network backflow to hidden fermion determinant states, Phys. Rev. B 110, 115124 (2024)

    2024

  96. [104]

    Chen, Large-scale simulation of deep neural quantum states, Ph.D

    A. Chen, Large-scale simulation of deep neural quantum states, Ph.D. thesis, Universit¨ at Augsburg (2025)

    2025

  97. [105]

    Roth and A

    C. Roth and A. H. MacDonald, Group convolutional neu- ral networks improve quantum state accuracy (2021), arXiv:2104.05085 [quant-ph]

    2021 arXiv

  98. [106]

    Cohen and M

    T. Cohen and M. Welling, Group equivariant convolu- tional networks, in Proceedings of The 33rd International Conference on Machine Learning , Proceedings of Ma- chine Learning Research, Vol. 48, edited by M. F. Balcan and K. Q. Weinberger (PMLR, New York, New York, USA, 2016)...

    2016

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.