Z-Error Loss masks high-z-score samples within each batch to reduce outlier influence in neural network training, but the paper provides no validation against baselines.
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
REJECT 1representative citing papers
citing papers explorer
-
Z-Error Loss for Training Neural Networks
Z-Error Loss masks high-z-score samples within each batch to reduce outlier influence in neural network training, but the paper provides no validation against baselines.