Tri-WE achieves state-of-the-art few-shot class-incremental learning by interpolating the classification heads of three consecutive models in weight space and distilling the previous model on CutMix-amplified data.
Incremental few-shot learn- ing via vector quantization in deep embedded space
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
-
Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning
Tri-WE achieves state-of-the-art few-shot class-incremental learning by interpolating the classification heads of three consecutive models in weight space and distilling the previous model on CutMix-amplified data.