betaFeedback

WorksComputer Science

Diagonal Symmetrization of Neural Network Solvers for the Many-Electron Schrödinger Equation

Abstract

Incorporating group symmetries into neural networks has been a cornerstone of success in many AI-for-science applications. Diagonal groups of isometries, which describe the invariance under a simultaneous movement of multiple objects, arise naturally in many-body quantum problems. Despite their importance, diagonal groups have received relatively little attention, as they lack a natural choice of invariant maps except in special cases. We study different ways of incorporating diagonal invariance in neural network ansätze trained via variational Monte Carlo methods, and consider specifically data augmentation, group averaging and canonicalization. We show that, contrary to standard ML setups, in-training symmetrization destabilizes training and can lead to worse performance. Our theoretical and numerical results indicate that this unexpected behavior may arise from a unique computational-statistical tradeoff not found in standard ML analyses of symmetrization. Meanwhile, we demonstrate that post hoc averaging is less sensitive to such tradeoffs and emerges as a simple, flexible and effective method for improving neural network solvers.

Introduction

We study the effect of symmetrizing neural network solutions to the Schrödinger equation. Solving the many-body Schrödinger equation is of fundamental importance in science, because it provides the key to understanding and predicting the behavior of quantum systems and thereby many physical phenomena. Ab initio computational methods seek to solve the non-relativistic electronic Schrödinger equation from first principles. There, the computation is performed directly from physical constraints and without relying on empirical approximations or training data, with the promise of producing high-accuracy electronic wavefunctions. However, the strict requirement on physical constraints makes it challenging to incorporate neural networks into these methods. Carleo & Troyer [9], Hermann et al. [22] and FermiNet Pfau et al., [40] are some of the first successful ab initio neural network methods, which learn the ground state wavefunctions in atoms and molecules via a variational Monte Carlo (VMC) approach. Many neural network methods have emerged since then [10] (Li et al. 2022; von Glehn et al. 2023), each seeking to improve how physical constraints are modelled in different systems. While these approaches have produced state-of-the-art results on ground state energy and other physical properties, one notable drawback is their exorbitant training cost compared to classical VMC methods. This issue becomes dire for modelling large systems, as the Hilbert space of wavefunctions grows exponentially with the number of electrons.

In other AI-for-science approaches, symmetrization has proved to be successful both for modelling physical constraints and for improving neural network performance [6, 16] (Batatıa et al. 2022; Du et al. 2022). One VMC example is DeepSolid Li et al., 2022, a FermiNet-type wavefunction that employs translationally invariant features to model periodic solids. However, recent findings in protein structures Abramson et al., [1], atomic potential Qu & Krishnapriyan, 2024 and machine learning theory Huang et al., [24]; Balestriero et al. [4] demonstrate that symmetries may be unnecessary and could sometimes be harmful for performance. This raises the question how much the previously observed improvements can be attributed to symmetries versus other factors such as hyperparameter and architecture choices.

Motivated by these questions, we examine the effectiveness of symmetrizing ab initio neural network solvers under the natural symmetry groups of a many-body problem, which are the diagonal groups of isometries. These groups, roughly speaking, describe the invariance of the system under a simultaneous movement of multiple objects; see § 2 for a detailed review. Common symmetrization approaches in machine learning (ML) roughly fall under three categories:

  • Randomly transforming data by symmetry operations (data augmentation);

  • Averaging over group operations (group averaging);

  • Invariant maps (invariant features and canonicalization).

We compare these approaches on DeepSolid (Li et al., 2022), the state-of-the-arts architecture for VMC on solids, for different solid systems. The emphasis is on providing an apples-to-apples comparison by fixing the architecture and hyperparameters, while varying the symmetry parameters. We find that, perhaps surprisingly, the effects of diagonal symmetrization for VMC problems are mixed and nuanced. This arises due to the unique combination of challenges posed by VMC and diagonal invariance, as well as the joint consideration of computational cost, statistical behaviors and physical constraints. In particular, our analyses indicate:

In-training symmetrization can hurt. VMC training operates

in an “infinite-data” regime, and every symmetry operation comes at the cost of forgoing one new data point. Holding the computational budget constant, symmetrization can destabilize training and lead to worse performance. This is demonstrated by theoretical and numerical results in § 4.

Post hoc symmetrization helps. At inference time, VMC solvers are less sensitive to computational costs, and allow for averaging over a moderate number of group operations (§ 5). We show that post hoc averaging (PA) leads to improved energy, variance and symmetry properties of the learned wavefunction (Figure 1, Table 2). In one case, post hoc averaged DeepSolid achieves performance close to DeepSolid trained with 10×10\times more computational budget (§ 7).

Visualizations of the partial diagonal invariance of an unsymmetrized wavefunction versus a symmetrized wavefunction

Figure 1. Visualizations of the (partial) diagonal invariance of an unsymmetrized wavefunction versus a symmetrized wavefunction in a graphene 1×11 \times1 system. (a) and (b) are generated by evaluating log⁡∣ψ(x1+t,…,x12+t)∣2\log\lvert\psi(x_1+t,\ldots,x_{12}+t)\rvert^2 under a simultaneous 2d translation tt of the configuration (x1,…,x12)(x_1,\ldots,x_{12}) given by the 12 blue spheres in (e). The red overlay indicates the unit cell in (e) such that the ref. line is exactly at the origin when t=0t=0. (d) shows 2 atoms (orange spheres) in an 1×11 \times1 planar supercell, with GG illustrated by the objects. (e) and (f) are shifted copies of the same configuration, with their positions marked in (b). This method only visualizes the partial P3m1 symmetry; see Appendix § 13 for a method that shows the full P6mm symmetry of ψθ(PA;G)\psi_\theta^{(\mathrm{PA;G})}. Details setups are in § 6, 7 and Appendix § 12.

SystemG\mathcal{G}MethodNNkkStepsGPU hours*Energy (Ha)Var⁡[Elocal]\operatorname{Var}[E_{\mathrm{local}}] (Ha2^2)Var[PA/OG]
Graphene 1 × 1-OG1000-80,000281−76.039(6)2.02(3)
DA901280,000277−76.039(3)2.48(5)
GA901280,000304−76.049(3)0.58(2)
GA10001210,000351−76.034(5)1.2(2)
GAs1000180,000281−76.049(3)0.48(2)
PA10001280,000281−76.050(3)0.33(1)0.00180(4)
PC10001280,000281−70.1(9)8(1)×1028(1) \times10^2
Lithium Hydride (LiH) 2 × 2 × 2P1ˉP\bar{1}OG-−8.138(2)0.06(1)
P1ˉP\bar{1}PA2−8.144(1)0.0344(9)0.0183(4)
P2/mP2/mPA4−8.148(1)0.0197(6)0.0249(5)
F222F222PA40001630,000571−8.1495(9)0.0162(7)0.0265(5)
Pm3ˉmPm\bar{3}mPA48−8.1502(7)0.0122(7)0.0289(8)
Fm3ˉmFm\bar{3}mPA192−8.1507(8)0.0118(7)0.0289(8)
Metallic Lithium (bcc-Li) 2 × 2 × 2-OG-−15.011(1)0.059(2)
P4/mmmP4/mmmPA30001620,000462−15.021(2)0.033(2)0.092(5)
FmmmFmmmPA32−15.020(1)0.036(2)0.0101(4)
Im3ˉmIm\bar{3}mPA96−15.022(3)0.031(3)0.0139(6)

Table 2. Performance of symmetrization methods with similar computational budgets. Energy and variance are both reported at the per unit cell level. *See Appendix § 12.§ 3 for specifications of the GPUs used for training.

The remainder of the paper provides mathematical discussions and computational tools for understanding diagonal symmetries in VMC. § 2 reviews the concept of diagonal invariance and discusses why, except for simple cases e.g., translations or E(3)E(3), finding a natural smooth invariant map is difficult. This is further corroborated by mathematical and empirical results in the special case of a smoothed canonicalization (§ 4.3 and Appendix § 15). § 3 briefly reviews the VMC setup and how different computational costs arise. § 4 and 5 respectively examine in-training and post hoc symmetrization. § 6 discusses our evaluation metrics and develops a method for visualizing diagonal symmetries (Figure 1). § 7 discusses experiment details. Additional results and proofs are included in the appendix.

Many-body Schrödinger equation

Throughout, we shall consider finding the nn-electron ground state wavefunction ψ:R3n→C\psi: \mathbb{R}^{3n} \to\mathbb{C} and the corresponding minimal energy E∈RE \in\mathbb{R} of the Schrödinger equation

Hψ(x)=Eψ(x),x:=(x1,…,xn)∈R3n.(1)H\psi(\mathbf{x}) = E\psi(\mathbf{x}), \quad\mathbf{x} := (x_1,\ldots,x_n) \in\mathbb{R}^{3n}. \tag*{(1)}

The Hamiltonian HH is given by Hψ(x):=−12Δψ(x)+V(x)ψ(x)H\psi(\mathbf{x}) := -\frac{1}{2}\Delta\psi(\mathbf{x}) + V(\mathbf{x})\psi(\mathbf{x}), Δ\Delta is the Laplacian representing the kinetic energy and V:R3n→RV : \mathbb{R}^{3n} \to\mathbb{R} is the potential energy of the physical system. We also denote neural network ansätze by ψθ\psi_\theta, parametrized by some network weights θ∈Rq\theta\in\mathbb{R}^q. Note that in general, the wavefunction depends on each electron via (xi,σi)(x_i,\sigma_i), where σi∈{↑,↓}\sigma_i \in\{\uparrow,\downarrow\} is the spin, and ψ\psi is required to be anti-symmetric with respect to permutations of (xi,σi)(x_i,\sigma_i). We focus on the case with fixed spins for simplicity, as is done in FermiNet and DeepSolid.

Diagonal invariance

Many physical systems of interest exhibit symmetry under some group G\mathbb{G} of isometries. G\mathbb{G} consists of maps of the form

x↦Ax+b,x∈R3,(2)x \mapsto Ax+b,\qquad x\in\mathbb{R}^3, \tag*{(2)}

for some orthogonal A∈R3×3A\in\mathbb{R}^{3\times3} and some translation b∈R3b\in\mathbb{R}^3. We focus on groups that are countable. For systems with n>1n>1 electrons, G\mathbb{G} typically causes the potential VV in (1) to be invariant under a diagonal group Gdiag\mathbb{G}_{\mathrm{diag}} acting on R3n\mathbb{R}^{3n}:

Gdiag:={(g,…,g)∣g∈G}.\mathbb{G}_{\mathrm{diag}} := \{(g,\ldots,g)\mid g\in\mathbb{G}\}.

To see how Gdiag\mathbb{G}_{\mathrm{diag}} may arise, consider the Coulomb potential

VCoul(x)=∑i<j1∥xi−xj∥+∑i,I1∥xi−rI∥+⋯ .V_{\mathrm{Coul}}(\mathbf{x})=\sum_{i<j}\frac{1}{\lVert x_i-x_j\rVert}+\sum_{i,I}\frac{1}{\lVert x_i-r_I\rVert}+\cdots.

Each rIr_I is the fixed, known position of the II-th atom (under the Born-Oppenheimer approximation), ∥∙∥\lVert\bullet\rVert is the Euclidean norm, and …\ldots are the omitted electron-independent terms.

terms. If the set of atom positions {rI}\{r_I\} is invariant under some G\mathbb{G}, VCoulV_{\mathrm{Coul}} is invariant under a simultaneous transformation of all xix_i’s by the same g∈Gg\in\mathbb{G}. Note however that VCoulV_{\mathrm{Coul}} does not satisfy separate invariance, i.e. invariance does not hold if x1x_1 and x2x_2 are transformed by different group elements, since

1∥g1(x1)−g2(x2)∥≠1∥x1−x2∥\frac{1}{\lVert g_1(x_1)-g_2(x_2)\rVert}\ne\frac{1}{\lVert x_1-x_2\rVert}

in general. Mathematically, separate invariance can be modelled by the product group Gn\mathbb{G}^n acting on R3n\mathbb{R}^{3n}, and diagonal invariance arises as a specific subgroup of Gn\mathbb{G}^n, i.e. Gdiag⊂Gn\mathbb{G}_{\mathrm{diag}}\subset\mathbb{G}^n. Figure 2 illustrates the difference between Gdiag\mathbb{G}_{\mathrm{diag}} and Gn\mathbb{G}^n in a 2-electron system: Under the given symmetry, a potential function is left unchanged when both electrons are reflected, but not when one is reflected and the other is rotated.

Different invariances for two electrons in a graphene system

Figure 2. Different invariances for two electrons in a graphene system. Left. Separate invariance under a reflection and a rotation. Right. Diagonal invariance under a simultaneous reflection.

Gdiag\mathbb{G}_{\mathrm{diag}}-invariant wavefunction. Throughout this paper, we use the shorthand g(x)=(g(x1),…,g(xn))g(\mathbf{x})=(g(x_1),\ldots,g(x_n)) for g∈Gg\in\mathbb{G}, and focus on Gdiag\mathbb{G}_{\mathrm{diag}}-invariant potentials VV, i.e.,

V(x)=V(g(x))for all g∈G.(3)V(\mathbf{x})=V(g(\mathbf{x}))\qquad\text{for all }g\in\mathbb{G}. \tag*{(3)}

Invariance of VV does not imply the invariance of ψ\psi: For a translation-invariant VV with one electron, Bloch [7] proves that ψ\psi is only invariant up to a phase factor, and non-invariant solutions can occur when the ground state is degenerate (Tinkham, [45]). Nonetheless, an invariant solution can always be constructed from a linear combination of these states. The next result confirms this for the general case of n≥1n\geq1 electrons and a diagonal group of isometries.

Fact 2.1. Suppose (ψ,E)(\psi,E) solves (1) and VV is invariant under some Gdiag\mathbb{G}_{\mathrm{diag}} induced by a group G\mathbb{G} of isometries. Then for any finite subgroup G\mathcal{G} of G\mathbb{G},

ψG(x):=1∣G∣∑g∈Gψ(g(x))(4)\psi^{\mathcal{G}}(\mathbf{x}):=\frac{1}{|\mathcal{G}|}\sum_{g\in\mathcal{G}}\psi(g(\mathbf{x})) \tag*{(4)}

also solves (1) with respect to same energy EE. In particular, if G\mathcal{G} is a subgroup, ψG\psi^{\mathcal{G}} is invariant under G\mathcal{G}.

Fact 2.1 motivates us to seek wavefunctions that respect the Gdiag\mathbb{G}_{\mathrm{diag}}-invariance of the system. Figure 1(a) also shows that an unsymmetrized, well-trained wavefunction already attempts to achieve some extent of approximate invariance.

For more references on how Gdiag\mathbb{G}_{\mathrm{diag}}-invariance arises and plays an important role in modelling ψ\psi, see Rajagopal et al. [42]; Zicovich-Wilson & Dovesi [51].

Challenges in modelling Gdiag\mathbb{G}_{\mathrm{diag}}-invariance. Despite successes in modelling simple nn-electron symmetries such as the Euclidean group E(3)E(3) (Batzner et al., [6]) and translations (Whitehead et al., [46]), generalizing the approaches in those settings to general isometries can be challenging:

  • Translation groups possess simple and well-understood symmetries. Common symmetrizations include periodic Fourier bases (Rajagopal et al., [42]) and projection via taking a modulus (Dym et al., [17]). The former trades off representation power against computational cost as the

number of bases used varies, whereas the latter suffers from discontinuity and requires smoothing (§ 4.3). Neither has been extended to a general GdiagG_{\mathrm{diag}}.

  • E(3)E(3) consists of all isometries in R3\mathbb{R}^{3}, and also admits simple invariant features ([6]). However, the additional symmetries under E(3)E(3) are undesirable when GG is only a subgroup of E(3)E(3), since the missing asymmetries represent a loss of information and therefore limit the representative power of the feature. Indeed, building an invariant map without losing information necessitates the availability of a maximal invariant ([30]), whereas building a maximal invariant for GdiagG_{\mathrm{diag}} while respecting continuity constraints requires extending the mathematical theory of orbifolds ([43, 2]) to GdiagG_{\mathrm{diag}} of isometries. Such extensions are not known to the best of our knowledge.

In short, invariant maps are well-understood for groups with simple isometries and with many isometries, and the difficult cases are often found in the ones with restricted symmetries. This problem is particularly pronounced in the case of a diagonal group. Mathematically, the symmetries of GdiagG_{\mathrm{diag}} are described by a fixed group GG in R3\mathbb{R}^{3}, but act on a larger and larger space R3n\mathbb{R}^{3n} as the number of electrons nn grows. This means that GdiagG_{\mathrm{diag}} admits substantially less structure than that of e.g. the product group GnG^{n}.

Space group GspG_{\mathrm{sp}} for a crystal system. A particular difficult case of GdiagG_{\mathrm{diag}}-invariance, in view of the above discussion, is the one induced by a space group GspG_{\mathrm{sp}} in a crystal lattice. These groups are described by (2) with a finite number of orthogonal matrices AA and a countable number of translations bb, which tile the R3\mathbb{R}^{3} space with a finite unit volume called the fundamental region. Figure 1(d) illustrates the P6mmP6mm point group in a graphene system, where the space is tiled by a triangular fundamental region. The study of space group is a fundamental subject of solid state physics, and we refer interested readers to [3].

Since GspG_{\mathrm{sp}} is infinite, Fact 2.1 does not apply directly to G=GspG=G_{\mathrm{sp}}. In practice however, to avoid computing an infinite crystal lattice, a common VMC practice is to restrict the system to a finite volume called the supercell ([18, 29]). This effectively reduces the set of translations bb in (2) to a finite set and hence GspG_{\mathrm{sp}} to a finite group, which allows Fact 2.1 to apply.

While our theoretical results apply to all finite groups of isometries, our numerical experiments focus on GspG_{\mathrm{sp}}. Each GspG_{\mathrm{sp}} is denoted by standard shorthand e.g., P6mmP6mm, and an exhaustive list of GspG_{\mathrm{sp}} can be found in Brock et al. [8]. We emphasize that our focus on the diagonal action of GspG_{\mathrm{sp}}, a difficult class of restricted symmetries found in crystals, sets our work apart from existing VMC works that have investigated translations, simple point symmetries in molecules, as well as the continuous symmetries in SO(2)SO(2), SU(2)SU(2) and E(3)E(3). A non-exhaustive list of those works can be found in [36, 33, 34, 50] and further comparisons are discussed in § 4.3.

Symmetrization of many-body wavefunctions. We shall focus on generic symmetrization techniques from ML that can be applied directly to many-electron wavefunctions. This is to be distinguished with classical techniques, e.g., [51]: There, symmetrization is performed by building invariance into single-electron features, which does not generalize well to state-of-the-art neural network solvers that are many-body by design.

Neural network VMC solver

We briefly review the VMC approach for training our ab initio neural network solvers. VMC seeks to solve the minimum eigenvalue problem of (1) via the optimization

References

  1. [1]Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A. J., Bambrick, J., et al. Accurate structure prediction of biomolecular interactions with alphafold 3. Nature, 630(8016):493–500, 2024.DOI
  2. [2]Adams, R. P. and Orbanz, P. Representing and learning functions invariant under crystallographic groups. arXiv preprint arXiv:2306.05261, 2023.arxiv.org/abs/2306.05261
  3. [3]Ashcroft, N. and Mermin, N. Solid State Physics. Cengage Learning, 1976.
  4. [4]Balestriero, R., Bottou, L., and LeCun, Y. The effects of regularization and data augmentation are class dependent. In Conference on Neural Information Processing Systems, 2022.arxiv.org/abs/2204.03632
  5. [6]Batzner, S., Musaelian, A., Sun, L., Geiger, M., Mailoa, J. P., Kornbluth, M., Molinari, N., Smidt, T. E., and Kozinsky, B. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nature communications, 13(1):2453, 2022.DOI
  6. [7]Bloch, F. Über die quantenmechanik der elektronen in kristallgittern. Zeitschrift für physik, 52(7):555–600, 1929.DOI
  7. [8]Brock, C. P., Hahn, T., Wondratschek, H., Müller, U., Shmueli, U., Prince, E., Authier, A., Kopský, V., Litvin, D., Arnold, E., et al. International tables for crystallography volume A: Space-group symmetry. Wiley Online Library, 2016.
  8. [9]Carleo, G. and Troyer, M. Solving the quantum many-body problem with artificial neural networks. Science, 355(6325):602–606, 2017.DOI
  9. [10]Cassella, G., Sutterud, H., Azadi, S., Drummond, N., Pfau, D., Spencer, J. S., and Foulkes, W. M. C. Discovering quantum phase transitions with fermionic neural networks. Physical Review Letters, 130(3):036401, 2023.DOI
  10. [11]Chen, L. H., Goldstein, L., and Shao, Q.-M. Normal approximation by Stein’s method, volume 2. Springer Science & Business Media, Berlin, 2011.
  11. [12]Chen, S., Dobriban, E., and Lee, J. H. A group-theoretic framework for data augmentation. Journal of Machine Learning Research, 21(245):1–71, 2020.arxiv.org/abs/1907.10905
  12. [13]Chernozhukov, V., Chetverikov, D., and Kato, K. Central limit theorems and bootstrap in high dimensions. Annals of Probability, 45(4):2309–2352, 2017.
  13. [16]Duval, A. A., Schmidt, V., Hernández-García, A., Miret, S., Malarrios, F. D., Bengio, Y., and Rolnick, D. FAENet: Frame averaging equivariant graph neural networks for materials modeling. In International Conference on Machine Learning, pp. 9013–9033. PMLR, 2023.
  14. [17]Dym, N., Lawrence, H., and Siegel, J. W. Equivariant frames and the impossibility of continuous canonicalization. arXiv preprint arXiv:2402.16077, 2024.arxiv.org/abs/2402.16077
  15. [18]Esler, K., Cohen, R., Militzer, B., Kim, J., Needs, R., and Towler, M. Fundamental high-pressure calibration from all-electron quantum monte carlo calculations. Physical Review Letters, 104(18):185702, 2010.
  16. [19]Gao, N. and Günnemann, S. Ab-initio potential energy surfaces by pairing gmns with neural wave functions. arXiv preprint arXiv:2110.05064, 2021.arxiv.org/abs/2110.05064
  17. [20]Gerard, L., Scherbela, M., Marquetand, P., and Grohs, P. Gold-standard solutions to the schrödinger equation using deep learning: How much physics do we need? Advances in Neural Information Processing Systems, 35:10282–10294, 2022.arxiv.org/abs/2205.09438
  18. [21]Han, J., Zhang, L., et al. Solving many-electron schrödinger equation using deep neural networks. Journal of Computational Physics, 399:108929, 2019.DOI
  19. [22]Hermann, J., Schätzle, Z., and Noé, F. Deep-neural-network solution of the electronic schrödinger equation. Nature Chemistry, 12(10):891–897, 2020.DOI
  20. [23]Hinton, G. E. Training products of experts by minimizing contrastive divergence. Neural computation, 14(8):1771–1800, 2002.
  21. [24]Huang, K. H., Orbanz, P., and Austern, M. Gaussian and non-gaussian universality of data augmentation. arXiv preprint arXiv:2202.09134, 1, 2022.arxiv.org/abs/2202.09134
  22. [25]Huang, K. H., Liu, X., Duncan, A., and Gandy, A. A high-dimensional convergence theorem for u-statistics with applications to kernel-based testing. In The Thirty Sixth Annual Conference on Learning Theory, pp. 3827–3918. PMLR, 2023.arxiv.org/abs/2302.05686
  23. [26]Jain, A., Ong, S. P., Hautier, G., Chen, W., Richards, W. D., Dacek, S., Cholia, S., Gunter, D., Skinner, D., Ceder, G., et al. Commentary: The materials project: A materials genome approach to accelerating materials innovation. APL materials, 1(1), 2013.DOI
  24. [27]Kaba, S.-O., Mondal, A. K., Zhang, Y., Bengio, Y., and Ravanbakhsh, S. Equivariance with learned canonicalization functions. In International Conference on Machine Learning, pp. 15546–15566. PMLR, 2023.
  25. [28]Kent, P. R., Needs, R., and Rajagopal, G. Monte carlo energy and variance-minimization techniques for optimizing many-body wave functions. Physical Review B, 59(19):12344, 1999.DOI
  26. [29]Kittel, C. and McEuen, P. Introduction to solid state physics. John Wiley & Sons, 2018.
  27. [30]Lehmann, E. L., Romano, J. P., and Casella, G. Testing statistical hypotheses, volume 3. Springer, 1986.
  28. [33]Lin, J., Goldshlager, G., and Lin, L. Explicitly antisym-metrized neural network layers for variational monte carlo simulation. Journal of Computational Physics, 474:111765, 2023.DOI
  29. [34]Luo, D., Chen, Z., Hu, K., Zhao, Z., Hur, V. M., and Clark, B. K. Gauge-invariant and anyonic-symmetric autoregressive neural network for quantum lattice models. Physical Review Research, 5(1):013216, 2023.DOI
  30. [36]Mahajan, A. and Sharma, S. Symmetry-projected jastrow mean-field wave function in variational monte carlo. The Journal of Physical Chemistry A, 123(17):3911–3921, 2019.DOI
  31. [37]Martens, J. and Grosse, R. Optimizing neural networks with kronecker-factored approximate curvature. In International conference on machine learning, pp. 2408–2417. PMLR, 2015.arxiv.org/abs/1503.05671
  32. [38]Müller, A. Integral probability metrics and their generating classes of functions. Advances in applied probability, 29(2):429–443, 1997.
  33. [39]Perdew, J. P., Burke, K., and Ernzerhof, M. Generalized gradient approximation made simple. Physical review letters, 77(18):3865, 1996.
  34. [40]Pfau, D., Spencer, J. S., Matthews, A. G., and Foulkes, W. M. C. Ab initio solution of the many-electron schrödinger equation with deep neural networks. Physical review research, 2(3):033429, 2020.DOI
  35. [42]Rajagopal, G., Needs, R., James, A., Kenny, S., and Foulkes, W. Variational and diffusion quantum monte carlo calculations at nonzero wave vectors: Theory and application to diamond-structure germanium. Physical Review B, 51(16):10591, 1995.
  36. [43]Ratcliffe, J. Foundations of hyperbolic manifolds. Graduate Texts in Mathematics/Springer-Verlag, 149, 1994.
  37. [44]Shevtsova, I. Optimization of the structure of the moment bounds for accuracy of normal approximation for the distributions of sums of independent random variables. PhD thesis, Moscow State University, 2013.
  38. [45]Tinkham, M. Group theory and quantum mechanics. Courier Corporation, 2003. von Glehn, I., Spencer, J. S., and Pfau, D. A self-attention ansatz for ab-initio quantum chemistry. In The Eleventh International Conference on Learning Representations, 2023.
  39. [46]Whitehead, T., Michael, M., and Conduit, G. Jastrow correlation factor for periodic systems. Physical Review B, 94 (3):035157, 2016.
  40. [47]Williams, K. T., Yao, Y., Li, J., Chen, L., Shi, H., Motta, M., Niu, C., Ray, U., Guo, S., Anderson, R. J., et al. Direct comparison of many-body methods for realistic electronic hamiltonians. Physical Review X, 10(1):011041, 2020.DOI
  41. [48]Yao, G., Xu, J., and Wang, X. Pseudopotential variational quantum monte carlo approach to bcc lithium. Physical Review B, 54(12):8393, 1996.
  42. [49]Zepeda-Núñez, L., Chen, Y., Zhang, J., Jia, W., Zhang, L., and Lin, L. Deep density: circumventing the kohn-sham equations via symmetry preserving neural networks. Journal of Computational Physics, 443:110523, 2021.
  43. [50]Zhang, Y., Jiang, B., and Guo, H. Schrödingernet: A universal neural network solver for the schrödinger equation. Journal of Chemical Theory and Computation, 21 (2):670–677, 2025.DOI
  44. [51]Zicovich-Wilson, C. and Dovesi, R. On the use of symmetry-adapted crystalline orbitals in scf-lcao periodic calculations. ii. implementation of the self-consistent-field scheme and examples. International journal of quantum chemistry, 67(5):311–320, 1998.doi.org/10.1002/(sici)1097-461x(1998)67:5<311::aid-qua4>3.0.co;2-y

Paper details

vlnc:W_kh6p5ehr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents