A Fisher information penalty added to the loss during batchwise training is claimed to correct covariate shift and improve accuracy, but the supporting derivation and experiments are incomplete.
Keel data-mining software tool: data set repository, integration of algorithms and experimental analysis framework
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
-
Causal Covariate Shift Correction using Fisher information penalty
A Fisher information penalty added to the loss during batchwise training is claimed to correct covariate shift and improve accuracy, but the supporting derivation and experiments are incomplete.