GRAFT selects diverse training samples per batch via fast MaxVol on low-rank features and dynamically tunes the sample count using gradient projection error, claiming near-full accuracy at reduced compute and emissions.
Glister: Generalization based data subset selection for efficient and robust learning
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
cs.LG 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
GRAFT: Gradient-Aware Fast MaxVol Technique for Dynamic Data Sampling
GRAFT selects diverse training samples per batch via fast MaxVol on low-rank features and dynamically tunes the sample count using gradient projection error, claiming near-full accuracy at reduced compute and emissions.