Multilevel Monte Carlo for quantum mechanics on a lattice

08/07/2020
by   Karl Jansen, et al.
0

Monte Carlo simulations of quantum field theories on a lattice become increasingly expensive as the continuum limit is approached since the cost per independent sample grows with a high power of the inverse lattice spacing. Simulations on fine lattices suffer from critical slowdown, the rapid growth of autocorrelations in the Markov chain with decreasing lattice spacing. This causes a strong increase in the number of lattice configurations that have to be generated to obtain statistically significant results. This paper discusses hierarchical sampling methods to tame this growth in autocorrelations. Combined with multilevel variance reduction, this significantly reduces the computational cost of simulations for given tolerances ϵ_disc on the discretisation error and ϵ_stat on the statistical error. For an observable with lattice errors of order α and an integrated autocorrelation time that grows like τ_int∝ a^-z, multilevel Monte Carlo (MLMC) can reduce the cost from 𝒪(ϵ_stat^-2ϵ_disc^-(1+z)/α) to 𝒪(ϵ_stat^-2|logϵ_disc|^2+ϵ_disc^-1/α). Even higher performance gains are expected for simulations of quantum field theories in D dimensions. The efficiency of the approach is demonstrated on two model systems, including a topological oscillator that is badly affected by critical slowdown due to freezing of the topological charge. On fine lattices, the new methods are orders of magnitude faster than standard sampling based on Hybrid Monte Carlo. For high resolutions, MLMC can be used to accelerate even the cluster algorithm for the topological oscillator. Performance is further improved through perturbative matching which guarantees efficient coupling of theories on the multilevel hierarchy.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
03/13/2020

Equivariant flow-based sampling for lattice gauge theory

We define a class of machine-learned flow-based sampling algorithms for ...
research
12/16/2022

Estimating truncation effects of quantum bosonic systems using sampling algorithms

To simulate bosons on a qubit- or qudit-based quantum computer, one has ...
research
01/21/2022

Stochastic normalizing flows as non-equilibrium transformations

Normalizing flows are a class of deep generative models that provide a p...
research
03/21/2022

Hierarchical autoregressive neural networks for statistical systems

It was recently proposed that neural networks could be used to approxima...
research
06/03/2021

Machine Learning and Variational Algorithms for Lattice Field Theory

In lattice quantum field theory studies, parameters defining the lattice...
research
11/28/2022

Deflated Multigrid Multilevel Monte Carlo

In lattice QCD, the trace of the inverse of the discretized Dirac operat...
research
03/03/2020

Towards Novel Insights in Lattice Field Theory with Explainable Machine Learning

Machine learning has the potential to aid our understanding of phase str...

Please sign up or login with your details

Forgot password? Click here to reset