Tuning coreset sampling parameters on validation data improves downstream F1 and balanced accuracy, and can beat full-data training in several settings.
Data-dependent coresets for compressing neural networks with applications to generalization bounds
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
-
Improving Model Classification by Optimizing the Training Dataset
Tuning coreset sampling parameters on validation data improves downstream F1 and balanced accuracy, and can beat full-data training in several settings.