Sculpting Margin Penalty: Intra-Task Adapter Merging and Classifier Calibration for Few-Shot Class-Incremental Learning
IEEE Transactions on Circuits and Systems for Video Technology · 2026-07 · 36 (7) · 9589-9605
DOI: 10.1109/tcsvt.2026.3670880 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-07. Verfasst von Liang Bai und 8 weiteren Autor:innen. Die Arbeit nennt 67 Literaturangaben. Gefördert von: National Science and Technology Major Project, National Natural Science Foundation of China.
Abstract (Originalsprache)
Real-world applications often face data privacy constraints and high acquisition costs, making the assumption of sufficient training data in incremental tasks unrealistic and leading to significant performance degradation in class-incremental learning. Forward-compatible learning, which prospectively prepares for future tasks during base task training, has emerged as a promising solution for Few-Shot Class-Incremental Learning (FSCIL). However, existing methods still struggle to balance base-class discriminability and new-class generalization. Moreover, limited access to original data during incremental tasks often results in ambiguous inter-class decision boundaries. To address these challenges, we propose SMP (Sculpting Margin Penalty), a novel FSCIL method that strategically integrates margin penalties at different stages within the parameter-efficient fine-tuning paradigm. Specifically, we introduce the Margin-aware Intra-task Adapter Merging (MIAM) mechanism for base task learning. MIAM trains two sets of low-rank adapters with distinct classification losses: one with a margin penalty to enhance base-class discriminability, and the other without margin constraints to promote generalization to future new classes. These adapters are then adaptively merged to improve forward compatibility. Furthermore, we propose a Margin Penalty-based Classifier Calibration (MPCC) strategy to alleviate decision boundary ambiguity during incremental tasks. Extensive experiments on CIFAR100, ImageNet-R, and CUB200 demonstrate that SMP achieves state-of-the-art performance in FSCIL while maintaining a better balance between base and new classes. Code will be publicly available at https://github.com/beiyan1911/SMP
Förderung
- National Science and Technology Major Project
- National Natural Science Foundation of China
Zitieren
APA
Bai, L., Song, H., Li, J., Lin, Y., Fan, J., Fu, T., Ai, D., Xiao, D., & Yang, J. (2026). Sculpting Margin Penalty: Intra-Task Adapter Merging and Classifier Calibration for Few-Shot Class-Incremental Learning. IEEE Transactions on Circuits and Systems for Video Technology, 36(7), 9589-9605. https://doi.org/10.1109/tcsvt.2026.3670880
Vancouver
Bai L, Song H, Li J, Lin Y, Fan J, Fu T, et al. Sculpting Margin Penalty: Intra-Task Adapter Merging and Classifier Calibration for Few-Shot Class-Incremental Learning. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(7):9589-9605. doi:10.1109/tcsvt.2026.3670880
BibTeX
@article{Bai2026,
author = {Bai, Liang and Song, Hong and Li, Jinfu and Lin, Yucong and Fan, Jingfan and Fu, Tianyu and Ai, Danni and Xiao, Deqiang and Yang, Jian},
title = {Sculpting Margin Penalty: Intra-Task Adapter Merging and Classifier Calibration for Few-Shot Class-Incremental Learning},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2026},
volume = {36},
number = {7},
pages = {9589--9605},
doi = {10.1109/tcsvt.2026.3670880}
}
EUVIMED