IP Library › Granted Patent US 12,248,861
Granted Patent B2
US 12,248,861 · App. 17/011,569 · Granted Mar 11, 2025

Antisymmetric neural networks

Inventors: David Benjamin Pfau (London, GB); James Spencer (London, GB); Alexander Graeme de Garis Matthews (London, GB)
Assignee: DeepMind Technologies Limited
G06N3/04G06F17/18
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,248,861
App. No.
17/011,569
Granted
Mar 11, 2025
Kind
B2
Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing inputs using antisymmetric neural networks.

Claims (81)

1. A method comprising:

obtaining input data characterizing a chemical system having a plurality of nuclei and a plurality of electrons, the input data comprising electron features for each of the plurality of electrons and pair features for each of a plurality of pairs of the plurality of electrons; and

processing the input data using an antisymmetric neural network having a plurality of intermediate layers, wherein:

the antisymmetric neural network is configured to process the input data to generate as output a predicted value of one or more properties of the chemical system, the one or more properties comprising one or more of a wavefunction of the chemical system or a ground state energy of the chemical system,

each of the plurality of intermediate layers of the antisymmetric neural network is configured to generate a respective layer output for the intermediate layer from a respective layer input to the intermediate layer by applying a respective permutation-equivariant function to the respective layer input,

the antisymmetric neural network is configured to generate the predicted value from the respective layer output of the last intermediate layer of the plurality of intermediate layers,

a respective layer output of a last intermediate layer of the plurality of layers comprises a respective output stream for each of the plurality of electrons,

generating the predicted value comprises:

generating, from the respective layer output of the last intermediate layer, a respective input for each of a plurality of pairs of determinants, each pair of determinants including (i) one determinant that operates on a respective spin up matrix that is generated from respective output streams for spin up electrons and has a respective row and a respective column for each of the spin up electrons and (ii) another determinant that operates on a respective spin down matrix generated from respective output streams for spin down electrons and has a respective row and a respective column for each of the spin down electrons, and wherein each pair of determinants of the plurality of determinants has respective spin down and spin up matrices that are different from each other pair of determinants of the plurality of determinants;

determining a respective output of each determinant in the plurality of pairs of determinants from the respective input for the determinant; and

determining the predicted value from the respective outputs of each of the determinants in the plurality of pairs of determinants,

and

the antisymmetric neural network has been trained on training data to optimize at least a variational energy objective.

2. The method of claim 1 , wherein, for an intermediate layer:

the respective layer input comprises a separate input stream for each of the electrons and for each of the pairs of electrons,

the layer output comprises a separate output stream for each of the electrons and for each of the pairs of electrons, and

the intermediate layer is configured to operate on each input stream with a corresponding stream sub-layer.

3. The method of claim 2 , wherein, for a first intermediate layer of the plurality of intermediate layers, the input stream for each of the electrons is the electron features for the electron and the input stream for each of the pairs of electrons is the pair features for the pair of electrons.

4. The method of claim 2 , wherein, for each of the plurality of intermediate layers, the particular stream sub-layer corresponding to each of the electrons is configured to:

receive the input stream for the corresponding electron from the respective layer input;

generate a combined input for the corresponding electron from the input streams from the respective layer input;

concatenate the input stream generated by the particular stream sub-layer with the combined input to generate a concatenated input stream for the corresponding electron; and

process the concatenated input stream to generate a concatenated output stream for the corresponding electron.

5. The method of claim 4 , wherein the concatenated output stream is the output stream for the corresponding electron.

6. The method of claim 4 , wherein, for at least one of the intermediate layers, the output stream for the corresponding electron includes the concatenated output stream and a residual output stream generated by the stream sub-layer for the corresponding electron in a preceding intermediate layer.

7. The method of claim 4 , wherein processing the concatenated input stream to generate a concatenated output stream comprises:

applying a linear transformation to the concatenated input stream to generate a transformed stream; and

applying a non-linear activation function to the transformed stream to generate the concatenated output stream.

8. The method of claim 4 , wherein generating a combined input for the corresponding electron comprises:

for each of a plurality of electron spins, computing an average of the input streams corresponding to electrons that have the electron spin; and

including the averages in the combined input.

9. The method of claim 8 , wherein generating a combined input for the corresponding electron comprises:

for each of the plurality of electron spins, computing an average of the input streams corresponding to pairs of electrons that include the corresponding electron and an electron that has the electron spin; and

including the averages in the combined input.

10. The method of claim 2 , wherein for each of the plurality of intermediate layers, the particular stream sub-layer corresponding to each of the pairs is configured to:

receive the input stream for the corresponding pair from the respective layer input for the intermediate layer; and

process the input stream to generate an initial output stream.

11. The method of claim 10 , wherein the initial output stream is the output stream for the corresponding pair.

12. The method of claim 10 , wherein, for at least one of the intermediate layers, the output stream for the corresponding pair includes the initial output stream and a residual output stream generated by the stream sub-layer for the corresponding pair in a preceding intermediate layer.

13. The method of claim 1 , wherein the electron features include a difference vector that measures a difference in positions of the electron and a nucleus within the system.

14. The method of claim 13 , wherein the electron features include a norm of the difference vector.

15. The method of claim 1 , wherein the electron features include a feature identifying a spin of the electron.

16. The method of claim 1 , wherein the pair features include a difference vector that measures a difference in positions between the electrons in the pair within the system.

17. The method of claim 16 , wherein the pair features include a norm of the difference vector.

18. The method of claim 1 , wherein the pair features include a feature identifying spins of the electron in the pair.

19. The method of claim 1 , wherein generating, from the layer output of the last intermediate layer, a respective input for each of a plurality of pairs of determinants comprises, for each pair of determinants:

applying a first final spin-dependent linear transformation to the respective output streams for the spin up electrons to generate a final transformed output for the spin up electrons;

generating the respective spin-up matrix for the determinant in the pair that operates on the spin up electrons based on applying a plurality of exponentially-decaying envelopes to some or all of the final transformed output for the spin up electrons, wherein applying the exponentially-decaying envelopes enforces a boundary condition that the wavefunction of the chemical system goes to zero away from the nuclei of the chemical system;

applying a second final spin-dependent linear transformation to the respective output streams for the spin down electrons to generate a final transformed output for the spin down electrons; and

generating the respective spin-down matrix for the other determinant in the pair that operates on the spin down electrons based on applying the plurality of exponentially-decaying envelopes to some or all of the final transformed output for the spin down electrons.

20. A system comprising one or more computers and one or more non-transitory storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations comprising:

obtaining input data characterizing a chemical system having a plurality of nuclei and a plurality of electrons, the input data comprising electron features for each of the plurality of electrons and pair features for each of a plurality of pairs of the plurality of electrons; and

processing the input data using an antisymmetric neural network having a plurality of intermediate layers, wherein:

the antisymmetric neural network is configured to process the input data to generate as output a predicted value of one or more properties of the chemical system, the one or more properties comprising one or more of a wavefunction of the chemical system or a ground state energy of the chemical system,

each of the plurality of intermediate layers of the antisymmetric neural network is configured to generate a respective layer output for the intermediate layer from a respective layer input to the intermediate layer by applying a respective permutation-equivariant function to the respective layer input,

the antisymmetric neural network is configured to generate the predicted value from the respective layer output of the last intermediate layer of the plurality of intermediate layers,

a respective layer output of a last intermediate layer of the plurality of layers comprises a respective output stream for each of the plurality of electrons,

generating the predicted value comprises:

generating, from the respective layer output of the last intermediate layer, a respective input for each of a plurality of pairs of determinants, each pair of determinants including (i) one determinant that operates on a respective spin up matrix that is generated from respective output streams for spin up electrons and has a respective row and a respective column for each of the spin up electrons and (ii) another determinant that operates on a respective spin down matrix generated from respective output streams for spin down electrons and has a respective row and a respective column for each of the spin down electrons, and wherein each pair of determinants of the plurality of determinants has respective spin down and spin up matrices that are different from each other pair of determinants of the plurality of determinants;

determining a respective output of each determinant in the plurality of pairs of determinants from the respective input for the determinant; and

determining the predicted value from the respective outputs of each of the determinants in the plurality of pairs of determinants,

and

the antisymmetric neural network has been trained on training data to optimize at least a variational energy objective.

21. One or more non-transitory computer-readable storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:

obtaining input data characterizing a chemical system having a plurality of nuclei and a plurality of electrons, the input data comprising electron features for each of the plurality of electrons and pair features for each of a plurality of pairs of the plurality of electrons; and

processing the input data using an antisymmetric neural network having a plurality of intermediate layers, wherein:

the antisymmetric neural network is configured to process the input data to generate as output a predicted value of one or more properties of the chemical system, the one or more properties comprising one or more of a wavefunction of the chemical system or a ground state energy of the chemical system,

each of the plurality of intermediate layers of the antisymmetric neural network is configured to generate a respective layer output for the intermediate layer from a respective layer input to the intermediate layer by applying a respective permutation-equivariant function to the respective layer input,

the antisymmetric neural network is configured to generate the predicted value from the respective layer output of the last intermediate layer of the plurality of intermediate layers,

a respective layer output of a last intermediate layer of the plurality of layers comprises a respective output stream for each of the plurality of electrons,

generating the predicted value comprises:

generating, from the respective layer output of the last intermediate layer, a respective input for each of a plurality of pairs of determinants, each pair of determinants including (i) one determinant that operates on a respective spin up matrix that is generated from respective output streams for spin up electrons and has a respective row and a respective column for each of the spin up electrons and (ii) another determinant that operates on a respective spin down matrix generated from respective output streams for spin down electrons and has a respective row and a respective column for each of the spin down electrons, and wherein each pair of determinants of the plurality of determinants has respective spin down and spin up matrices that are different from each other pair of determinants of the plurality of determinants;

determining a respective output of each determinant in the plurality of pairs of determinants from the respective input for the determinant; and

determining the predicted value from the respective outputs of each of the determinants in the plurality of pairs of determinants,

and

the antisymmetric neural network has been trained on training data to optimize at least a variational energy objective.

22. The system of claim 20 , wherein, for an intermediate layer:

the respective layer input comprises a separate input stream for each of the electrons and for each of the pairs of electrons,

the layer output comprises a separate output stream for each of the electrons and for each of the pairs of electrons, and

the intermediate layer is configured to operate on each input stream with a corresponding stream sub-layer.

23. The system of claim 22 , wherein, for a first intermediate layer of the plurality of intermediate layers, the input stream for each of the electrons is the electron features for the electron and the input stream for each of the pairs of electrons is the pair features for the pair of electrons.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 6, 2025
From: DEEPMIND TECHNOLOGIES LIMITED
To: GDM HOLDING LLC
Reel/Frame 071498/0210 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 12, 2020
From: PFAU, DAVID BENJAMIN; SPENCER, JAMES; MATTHEWS, ALEXANDER GRAEME DE GARIS
To: DEEPMIND TECHNOLOGIES LIMITED
Reel/Frame 054030/0449 →
Continuity (2)
Provisional Application 62895491 · Sep 3, 2019
Related Publication 20210064961A1 · Mar 4, 2021
References Cited (88)
US 7788196B2 · Buscema · 2010 [cited by examiner]
US 20210398621A1 · Stojevic · 2021 [cited by examiner]
US 20220383992A1 · Triendl · 2022 [cited by examiner]
Han, Jiequn, Linfeng Zhang, and E. Weinan. “Solving many-electron Schrödinger equation using deep neural networks.” (2018) (Year: 2018). [cited by examiner]
Bond, Benjamin Daniel. “A Comparative Analysis of Partial Least Squares (PLS) Regression and Artificial Neural Network (ANN) Methods for Determining Elemental Weights in Sheet Metal” (2014). (Year: 2014). [cited by examiner]
Chang, Bo, et al. “Reversible architectures for arbitrarily deep residual neural networks.” Proceedings of the AAAI conference on artificial intelligence. vol. 32. No. 1. 2018. (Year: 2018). [cited by examiner]
Unke, Oliver T., and Markus Meuwly. “PhysNet: A neural network for predicting energies, forces, dipole moments, and partial charges.” Journal of chemical theory and computation 15.6 (2019): 3678-3693. (Year: 2019). [cited by examiner]
Zhai, Huanchen. Potential Energy Surface Exploration of Metal Catalytic Clusters. University of California, Los Angeles, 2019. (Year: 2019). [cited by examiner]
Han, Jiequn, et al. “Universal approximation of symmetric and anti-symmetric functions.” arXiv preprint arXiv:1912.01765 (2019). (Year: 2019). [cited by examiner]
Lo, Shih Chung B., et al. “Transformationally identical and invariant convolutional neural networks through symmetric element operators.” arXiv preprint arXiv:1806.03636 (2018). (Year: 2018). [cited by examiner]
Ognjanovski G. (2019). Everything you need to know about neural networks and backpropagation—Machine learning easy and fun. https://towardsdatascience.com/everything-you-need-to-know-about-neural-networks-and-backpropag… [cited by examiner]
International Preliminary Report on Patentability in International Appln. No. PCT/EP2020/074578, dated Mar. 8, 2022, 9 pages. [cited by applicant]
Amari, “Natural gradient works efficiently in learning,” Neural Comput., Feb. 1998, 10(2):251-76. [cited by applicant]
Assaraf et al., “Optimizing the energy with quantum Monte Carlo: A lower numerical scaling for Jastrow-Slater expansions,” Journal of chemical theory and computation, Nov. 2017, 14;13(11):5273-81. [cited by applicant]
Badinski et al., “Methods for calculating forces within quantum Monte Carlo simulations,” Journal of Physics: Condensed Matter, Feb. 2010, 22(7):074202. [cited by applicant]
Bajdich et al., “We will list here the peer-reviewed publications that have used QWalk in one form or another,” Phys. Rev. Lett., 2006, 96:130201. [cited by applicant]
Bartlett et al., “Coupled-cluster theory in quantum chemistry,” Rev. Modern Phys., Feb. 2007, 79(1):291-352. [cited by applicant]
Booth et al., “Approaching chemical accuracy using full configuration interaction quantum Monte Carlo: a study of ionization potentials,” J. Chem. Phys., May 2010, 132(17):174104. [cited by applicant]
Burton et al., “Holomorphic Hartree-Fock Theory: An Inherently Multireference Approach,” J. Chem. Theory Comput., Jan. 2016, 12(1):167-73. [cited by applicant]
Carleo et al., “Solving the quantummany-body problemwith artificial neural networks,” Science, Feb. 2017, 355(6325):602-6. [cited by applicant]
Cencek et al., “Benchmark calculations for He2+ and LiH molecules using explicitly correlated gaussian functions,” Chem. Phys. Lett., Apr. 2000, 320(5-6):549-552. [cited by applicant]
Chakravorty et al., “Ground-state correlation energies for atomic ions with 3 to 18 electrons,” Phys. Rev. A, May 1993, 47(5):3649. [cited by applicant]
Cheng et al., “A universal density matrix functional from molecular orbital-based machine learning: Transferability across organic molecules,” J. Chem. Phys., 2019, 150:131103. [cited by applicant]
Choo et al., “Fermionic neural-network states for ab-initio electronic structure,” Nature communications, May 2020, 11(1):1-7. [cited by applicant]
Choo et al., “Symmetries and many-body excita262 tions with neural-network quantum states,” Phys. Rev. Lett., Oct. 2018, 121(16):167204. [cited by applicant]
Clark et al., “Computing the energy of a water molecule using multideterminants: A simple, efficient algorithm,” The Journal of chemical physics, Dec. 2011, 135(24):244105. [cited by applicant]
Curtiss et al., “Gaussian-3 (G3) theory for molecules containing first and second-row atoms,” J. Chem. Phys., Nov. 1998, 109(18):7764-7776. [cited by applicant]
Dash et al., “Perturbatively selected configuration-interaction wave functions for efficient geometry optimization in quan tum Monte Carlo,” Journal of chemical theory and computation, Jun. 2018, 14(8):4176-82. [cited by applicant]
DePrince et al., “Accuracy and efficiency of coupled-cluster theory using density fitting / Cholesky decomposition, frozen natural orbitals, and a T1-transformed Hamiltonian,” J. Chem. Theory Comput., Jun. 2013, 9(6):26… [cited by applicant]
Feller et al., “Application of systematic sequences of wave functions to the water dimer,” J. Chem. Phys., Apr. 1992, 96(8):6104-14. [cited by applicant]
Feynman et al., “Energy spectrum of the excitations in liquid helium,” Physical Review, Jun. 1956, 102(5):1189. [cited by applicant]
Filippi et al., “Multiconfiguration wave functions for quantum Monte Carlo calculations of first-row diatomic molecules,” J. Chem. Phys., Jul. 1996, 105(1):213-226. [cited by applicant]
Flyvbjerg et al., “Error estimates on averages of correlated data,” J. Chem. Phys., Jul. 1989, 91(1):461-466. [cited by applicant]
Foulkes et al., “Quantum Monte Carlo simulations of solids,” Rev. Modern Phys., Jan. 2001, 73(1):33. [cited by applicant]
Gdanitz et al., “Erratum to Accurately solving the electronic Schrodinger equation of atoms and molecules using explicitly correlated (r 1 2-) MR-CI: the ground state potential energy curve of N 2,” Chemical Physics Let… [cited by applicant]
Gdanitz, “Accurately solving the electronic Schrodinger equation of atoms and molecules using explicitly correlated (r12-)MR-CI: the ground state potential energy curve of N2,” Chem. Phys. Lett., Feb. 1998, 283(5-6):253… [cited by applicant]
Giles et al., “Collected matrix derivative results for forward and reverse mode algorithmic differentiation,” In Advances in Automatic Differentiation, 2008, 35-44. [cited by applicant]
Gilmer et al., “Neural Message Passing for Quantum Chemistry,” Proceedings of the 34th International Conference on Machine Learning, 2017, 0:1263-1272. [cited by applicant]
Giner et al., “Quantum Monte Carlo with reoptimised perturbatively selected configuration-interaction wave functions,” Molecular Physics, Apr. 2016, 114(7-8):910-20. [cited by applicant]
Han et al., “Solving many-electron Schrodinger equation using deep neural networks,” arXiv preprint arXiv:1807.07014, Dec. 2019, 399:108929. [cited by applicant]
Helgaker et al., “Basis-set convergence of correlated calculations on water,” J. Chem. Phys., Jun. 1997, 106(23):9639-46. [cited by applicant]
Hermann et al., “Deep-neural-network solution of the electronic Schrödinger equation,” Nature Chemistry, Oct. 2020, 12(10):891-7. [cited by applicant]
Hood et al., “Quantum monte carlo investigation of exchange and correlation in silicon,” Physical review letters, Apr. 1997, 78(17):3350. [cited by applicant]
Kessler et al., “Artificial neural networks as trial wave 268 functions for quantum Monte Carlo,” arXiv preprint arXiv:1904.10251, Apr. 2019, 14 pages. [cited by applicant]
Klopper et al., “Sub-meV accuracy in first286 principles computations of the ionization potentials and electron affinities of the atoms H to Ne,” Physical Review A., Feb. 2010, 81(2):022503. [cited by applicant]
Krizhevsky et al., “Imagenet classification with deep convolutional neural networks,” NeurIPS, May 2012, 60(6):84-90. [cited by applicant]
Le Roy et al., “An accurate analytic potential function for ground-state N2 from a direct-potential-fit analysis of spectroscopic data,” J. Chem. Phys., Oct. 2006, 125(16):164310. [cited by applicant]
Lee et al., “Strategies for improving the efficiency of quantum Monte Carlo calculations,” Physical Review E., Jun. 2011, 83(6):066706. [cited by applicant]
Lou et al., “Backflow Transformations via Neural Networks for Quantum Many-Body Wave-Functions,” CoRR, Jul. 2018, arxiv.org/abs/1807.10770, 10 pages. [cited by applicant]
Luo et al., “Machine learning many-electron wave functions via backflow transformations,” Phys. Rev. Lett., 2019, 122:226401. [cited by applicant]
Lyakh et al., “Multireference nature of chemistry: the coupled-cluster view,” Chem. Rev., Jan. 2012, 112(1):182-243. [cited by applicant]
Martens et al., “Optimizing neural networks with Kronecker-factored ap274 proximate curvature,” InInternational conference on machine learning, Jun. 2015, 2408-2417. [cited by applicant]
Mazzola et al., “Finite-temperature electronic simulations without the Born-Oppenheimer constraint,” The Journal of chemical physics, Oct. 2012, 137(13):134112. [cited by applicant]
Mills et al., “Deep learning and the Schrödinger equation,” Physical Review A., Oct. 2017, 96(4):042113. [cited by applicant]
Motta et al., “Ground-state properties of the hydrogen chain: insulator-to-metal transition, dimerization, and magnetic phases,” arXiv preprint arXiv:1911.01618, Nov. 2019. [cited by applicant]
Motta et al., “Towards the solution of the many-electron problem in real materials: equation of state of the hydrogen chain with state-of-the-artmany-bodymethods,” Phys. Rev. X, Sep. 2017, 7(3):031059. [cited by applicant]
Nagy et al., “Variational quantum Monte Carlo method with a neural264 network Ansatz for open quantum systems,” Phys. Rev. Lett., Jun. 2019, 122(25):250501. [cited by applicant]
Needs et al., “Casino: User's Guide Version 2.13,” Sep. 2019, pp. 180-181. [cited by applicant]
Neuscamman et al., “Optimizing large parameter sets in variational quantum Monte Carlo,” Physical Review B., Jan. 2012, 85(4):045103. [cited by applicant]
Nomura et al., “Restricted Boltzmann machine learning for solving strongly correlated quantum systems,” Physical Review B., Nov. 2017, 96(20):205152. [cited by applicant]
Orús et al., “A practical introduction to tensor networks: Matrix product states and projected entangled pair states,” Annals of Physics, Oct. 2014, 349:117-58. [cited by applicant]
Otis et al., “Complementary first and second derivative methods 278 for Ansatz optimization in variational Monte Carlo,” Phys. Chem. Chem. Phys., 2019, 21(27):14491-510. [cited by applicant]
Parrish et al., “Psi4 1.1: an open-source electronic structure program emphasizing automation, advanced libraries, and interoperability,” J. Chem. Theory Comput., Jul. 2017, 13(7):3185-3197. [cited by applicant]
PCT International Search Report and Written Opinion in International Appln. No. PCT/EP2020/074578, dated Dec. 14, 2020, 15 pages. [cited by applicant]
Petz et al., “Geometries of quantum states,” J. Math. Phys., Jun. 1996, 37(6):2662-2673. [cited by applicant]
Pfau et al., “Ab initio solution of the many-electron Schrödinger equation with deep neural networks,” Physical Review Research, Sep. 2020, 2(3):033429. [cited by applicant]
Ramachandran et al., “Stand-Alone Self-Attention in Vision Models,” 33rd Conference on Neural Information Processing Systems, 2019, 15 pages. [cited by applicant]
Ruggeri et al., “Nonlinear network description for many-body quantum systems in continuous space,” Physical review letters, May 2018, 120(20):205302. [cited by applicant]
Sabzevari et al., “An accelerated linear method for optimizing non-linear wavefunctions in variational Monte Carlo,” The Journal of Chemical Physics, Jan. 2020, 152(2):024111. [cited by applicant]
Saito et al., “Method to solve quantum few-body problems with artificial neural networks,” J. Phys. Soc. Japan, Jul. 2018, 87(7):074002. [cited by applicant]
Schütt et al., “ Quantum-chemical insights from deep tensor neural networks,” Nature communications, Jan. 2017, 8(1):1-8. [cited by applicant]
Schütt et al., “SchNet—A deep learning architecture for molecules and materials,” The Journal of Chemical Physics, Jun. 2018, 28;148(24):241722. [cited by applicant]
Schütt et al., “Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions,” Nature Communications, Nov. 2019, 10(1):1-0. [cited by applicant]
Senior et al., “Improved protein structure prediction using potentials from deep learning,” Nature, Jan. 2020, 577(7792):706-10. [cited by applicant]
Seth et al., “QuantumMonte Carlo study of the first-row atoms and ions,” J. Chem. Phys., Feb. 2011, 134(8):084105. [cited by applicant]
Shawe-Taylor et al., “Building symmetries into feedforward networks,” In1989 First IEE International Conference on Artificial Neural Networks, Oct. 1989, 158-162. [cited by applicant]
Silver et al., “Mastering the game of Go with deep neural networks and tree search,” Nature, Jan. 2016, 529(7587):484-9. [cited by applicant]
Sinitskiy et al., “Physical machine learning outperforms human learning in Quantum Chemistry,” arXiv preprint arXiv, Aug. 2019, 1908.00971. [cited by applicant]
Sorella et al., “Green function Monte Carlo with stochastic reconfiguration,” Phys. Rev. Lett., May 1998, 80(20):4558. [cited by applicant]
Sun et al., “PySCF: the Python-based simulations of chemistry framework,” Wiley Interdisciplinary Reviews: Computational Molecular Science, Jan. 2018, 8(1):e1340. [cited by applicant]
Taddei et al., “Iterative backflow renormalization procedure for many-body ground-state wave functions of strongly interacting normal Fermi liquids,” Physical Review B, Mar. 2015, 91(11):115106. [cited by applicant]
Toulouse et al., “Optimization of quantum Monte Carlo wave functions by energy minimization,” J. Chem. Phys., Feb. 2007, 126(8):084102. [cited by applicant]
Troyer et al., “Computational complexity and fundamental limitations to 245 fermionic quantum Monte Carlo simulations,” Physical review letters, May 2005, 94(17):170201. [cited by applicant]
Umrigar et al., “A diffusion Monte Carlo algorithm with very small time-step errors,” J. Chem. Phys., Aug. 1993, 99(4):2865-2890. [cited by applicant]
Van Voorhis et al., “Benchmark variational coupled cluster doubles results,” J. Chem. Phys., Nov. 2000, 113(20):8873-8879. [cited by applicant]
Yang et al., “Deep learning-enhanced variational Monte Carlo method for quantum many-body physics,” Physical Review Research, Feb. 2020, 2(1):012039. [cited by applicant]
Zhang et al., “Ab initio electronic structure calculations by auxiliary-field quantum Monte Carlo,” Handbook of Materials Modeling: Methods: Theory and Modeling, 2020, 123-49. [cited by applicant]
Decision to Grant Patent in Japanese Appln. No. 2022-511314, dated Jul. 3, 2023, 5 pages. [cited by applicant]