RegMixMatch combines clean-plus-mixed training with class-aware Mixup for low-confidence unlabeled samples, achieving state-of-the-art semi-supervised image classification.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learning
RegMixMatch combines clean-plus-mixed training with class-aware Mixup for low-confidence unlabeled samples, achieving state-of-the-art semi-supervised image classification.