Scaling Up Machine Learning For Quantum Field Theory with Equivariant Continuous Flows

de Haan, Pim; Rainone, Corrado; Cheng, Miranda C.N.; Bondesan, Roberto

Scaling Up Machine Learning For Quantum Field Theory with Equivariant Continuous Flows

Pim de Haan
(
- Qualcomm, Amsterdam
)
,
Corrado Rainone
(
- Qualcomm, Amsterdam
)
,
Miranda C.N. Cheng
(
- Amsterdam U., Inst. Math. and
- Academia Sinica, Taiwan
)
,
Roberto Bondesan
(
- Qualcomm, Amsterdam
)

Oct 6, 2021

8 pages

e-Print:

2110.02673 [cs.LG]

View in:

ADS Abstract Service

pdf

reference search34 citations

Citations per year

Abstract: (arXiv)

We propose a continuous normalizing flow for sampling from the high-dimensional probability distributions of Quantum Field Theories in Physics. In contrast to the deep architectures used so far for this task, our proposal is based on a shallow design and incorporates the symmetries of the problem. We test our model on the

\phi^4

theory, showing that it systematically outperforms a realNVP baseline in sampling efficiency, with the difference between the two increasing for larger lattices. On the largest lattice we consider, of size

32\times 32

, we improve a key metric, the effective sample size, from 1% to 66% w.r.t. the realNVP baseline.

Note:

8 pages, 5 figures. Fourth Workshop on Machine Learning and the Physical Sciences (NeurIPS 2021)

flow
lattice
scaling
efficiency

References(25)

Figures(5)

Seminars

[1]

Machine learning and the physical sciences

Giuseppe Carleo
(
- Flatiron Inst., New York
)
,
Ignacio Cirac
(
- Munich, Max Planck Inst. Quantenopt.
)
,
Kyle Cranmer
(
- New York U., CCPP
)
,
Laurent Daudet
(
- Unlisted, FR
)
,
Maria Schuld
(
- KwaZulu Natal U. and
- Xanadu Quant. Tech.
)

et al.

- Rev.Mod.Phys. 91 (2019) 4, 045002
•
e-Print:
- 1903.10563
•
DOI:
- 10.1103/RevModPhys.91.045002

[2]

Frank Noé, Alexandre Tkatchenko, Klaus-Robert Müller, and Cecilia Clementi. Machine learning for molecular simulation

- Ann.Rev.Phys.Chem. 71 (2020) 361-390

[3]

Machine learning at the energy and intensity frontiers of particle physics

Alexander Radovic
(
- William-Mary Coll.
)
,
Mike Williams
(
- MIT
)
,
David Rousseau
(
- Orsay, LAL
)
,
Michael Kagan
(
- SLAC
)
,
Daniele Bonacorsi
(
- U. Bologna (main) and
- INFN, Bologna
)

et al.

- Nature 560 (2018) 7716, 41-48
•
DOI:
- 10.1038/s41586-018-0361-2

[4]

Hsin-Yuan Huang, Richard Kueng, Giacomo Torlai

Victor V. Albert

[5]

Christian Robert and George Casella. Monte Carlo statistical methods

[6]

Flow-based generative models for Markov chain Monte Carlo in lattice field theory

M.S. Albergo
(
- Perimeter Inst. Theor. Phys. and
- Cambridge U. and
- Waterloo U.
)
,
G. Kanwar
(
- MIT, Cambridge, CTP
)
,
P.E. Shanahan
(
- MIT, Cambridge, CTP and
- Perimeter Inst. Theor. Phys.
)

- Phys.Rev.D 100 (2019) 3, 034515
•
e-Print:
- 1904.12072
•
DOI:
- 10.1103/PhysRevD.100.034515

[7]

Asymptotically unbiased estimation of physical observables with neural samplers

et al.

- Phys.Rev.E 101 (2020) 2, 023304,
- Phys. Rev. E, Volume 101, Issue 2, 02/2020
•
e-Print:
- 1910.13496
•
DOI:
- 10.1103/PhysRevE.101.023304

[8]

Reducing Autocorrelation Times in Lattice Simulations with Generative Adversarial Networks

- Mach.Learn.Sci.Tech. 1 (2020) 045011
•
e-Print:
- 1811.03533
•
DOI:
- 10.1088/2632-2153/abae73

[9]

Equivariant flow-based sampling for lattice gauge theory

Gurtej Kanwar
(
- MIT
)
,
Michael S. Albergo
(
- New York U., CCPP
)
,
Denis Boyda
(
- MIT
)
,
Kyle Cranmer
(
- New York U., CCPP
)
,
Daniel C. Hackett
(
- MIT
)

et al.

- Phys.Rev.Lett. 125 (2020) 12, 121601
•
e-Print:
- 2003.06413
•
DOI:
- 10.1103/PhysRevLett.125.121601

[10]

Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Julian M. Urban

Michael S. Albergo

[11]

Estimation of Thermodynamic Observables in Lattice Field Theories with Deep Generative Models

et al.

- Phys.Rev.Lett. 126 (2021) 3, 032001
•
e-Print:
- 2007.07115
•
DOI:
- 10.1103/PhysRevLett.126.032001

[12]

Luigi Del Debbio, Joe Marsh Rossney, and Michael Wilson. Efficient modelling of trivializing maps for lattice φ4 theory using normalizing flows: A first look at scalability

[13]

Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. Density estimation using real nvp

[14]

Frank Noé, Simon Olsson, Jonas Köhler, and Hao Wu. Boltzmann generators: Sampling equilibrium states of many-body systems with deep learning. Science, 365(6457)

- Science (2019)

[15]

Jonas Köhler, Leon Klein, and Frank Noe. Equivariant flows: Exact likelihood generative learning for symmetric densities. In Hal Daumé III and Aarti Singh, editors, Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 5361-5370. PMLR, 13-18 Jul

[16]

Akio Tomiya and Yuki Nagai. Gauge covariant neural network for 4 dimensional non-abelian gauge theory

[17]

István Montvay and Gernot Münster. Quantum fields on a lattice. Cambridge University Press

[18]

Tian Qi Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations. In Advances in neural information processing systems, pages 6571-6583

[19]

Chung-Chun Hsieh, Michael S. Albergo

Daniel C. Hackett

[20]

Matthew Tancik

Pratul P. Srinivasan

•
e-Print:
- 2006.10739

[21]

Denis Boyda, Daniel C. Hackett

Michael S. Albergo

[22]

Glow: Generative Flow with Invertible 1x1 Convolutions

e-Print:
- 1807.03039

[23]

Adam: A Method for Stochastic Optimization

Diederik P. Kingma
(
- Amsterdam U.
)
,
Jimmy Ba
(
- Toronto U.
)

- Published as a conference paper at the 3rd International Conference for Learning Representations, San Diego, 2015
•
e-Print:
- 1412.6980

[24]

Algorithm 1 Independent Metropolis Hastings algorithm. Require: q(φ): proposal density, p(φ): target density. for i = 1 : K do φ0 ∼ q(φ) u ∼ Uniform(u

I. Vierhaus

[24]

Simulation of ϕ4 theory in the strong coupling expansion beyond the Ising limit. PhD thesis, Humboldt University

A. Background Details A.1 Algorithms