The paper proposes two relevance functions and two sampling methods for imbalanced regression, and reports that crbSMOGN with density-ratio relevance improves rare-sample prediction for neural networks, while an ensemble with an unmitigated model limits the loss on frequent samples.
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
-
Model-agnostic Mitigation Strategies of Data Imbalance for Regression
The paper proposes two relevance functions and two sampling methods for imbalanced regression, and reports that crbSMOGN with density-ratio relevance improves rare-sample prediction for neural networks, while an ensemble with an unmitigated model limits the loss on frequent samples.