SAIL trains ML models for DFT initial guesses by backpropagating through the SCF solver, yielding transferable speedups that reduce ERIC by up to 37% on molecules 4-10x larger than training data.
To evaluate out-of-distribution generalization beyond QM9, we additionally use QM40 and QMugs as far out-of-distribution test sets
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Transferable SCF-Acceleration through Solver-Aligned Initialization Learning
SAIL trains ML models for DFT initial guesses by backpropagating through the SCF solver, yielding transferable speedups that reduce ERIC by up to 37% on molecules 4-10x larger than training data.