Training a surrogate model to minimize prediction error on the points visited by a downstream scientific-computing algorithm, rather than on the whole training distribution, can substantially improve the algorithm's output.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Learning task-specific predictive models for scientific computing
Training a surrogate model to minimize prediction error on the points visited by a downstream scientific-computing algorithm, rather than on the whole training distribution, can substantially improve the algorithm's output.