Fractional Classification Loss uses a learnable fractional derivative order to interpolate between cross-entropy-like and MAE-like behavior, adapting its robustness to label noise during training.
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
-
Introducing Fractional Classification Loss for Robust Learning with Noisy Labels
Fractional Classification Loss uses a learnable fractional derivative order to interpolate between cross-entropy-like and MAE-like behavior, adapting its robustness to label noise during training.