Neural-network control variates built from Schwinger-Dyson identities reduce variance in 1+1D scalar and 2D U(1) lattice theories, and a scalar-only construction is shown to be non-universal for a two-plaquette Wilson loop.
Cybenko,Approximation by superpositions of a sigmoidal function, Math
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
hep-lat 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Control variates with neural networks
Neural-network control variates built from Schwinger-Dyson identities reduce variance in 1+1D scalar and 2D U(1) lattice theories, and a scalar-only construction is shown to be non-universal for a two-plaquette Wilson loop.