A curriculum of nearest-neighbor transfer-learning steps, implemented with the Pairing-Net neural wavefunction, reduces the number of epochs needed to explore Hubbard-model parameter regimes by up to two orders of magnitude versus from-scratch variational Monte Carlo.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.str-el 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Explorative Curriculum Learning for Strongly Correlated Electron Systems
A curriculum of nearest-neighbor transfer-learning steps, implemented with the Pairing-Net neural wavefunction, reduces the number of epochs needed to explore Hubbard-model parameter regimes by up to two orders of magnitude versus from-scratch variational Monte Carlo.