Distilled Meta-learning for Multi-Class Incremental Learning
ACM Transactions on Multimedia Computing, Communications, and Applications · 31.07.2023 · 19 (4) · 1-16
DOI: 10.1145/3576045 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Multimedia Computing, Communications, and Applications, online erschienen am 15.03.2023. Verfasst von Hao Liu und 5 weiteren Autor:innen. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 45 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Jiangsu Province.
Abstract (Originalsprache)
Meta-learning approaches have recently achieved promising performance in multi-class incremental learning. However, meta-learners still suffer from catastrophic forgetting, i.e., they tend to forget the learned knowledge from the old tasks when they focus on rapidly adapting to the new classes of the current task. To solve this problem, we propose a novel distilled meta-learning (DML) framework for multi-class incremental learning that integrates seamlessly meta-learning with knowledge distillation in each incremental stage. Specifically, during inner-loop training, knowledge distillation is incorporated into the DML to overcome catastrophic forgetting. During outer-loop training, a meta-update rule is designed for the meta-learner to learn across tasks and quickly adapt to new tasks. By virtue of the bilevel optimization, our model is encouraged to reach a balance between the retention of old knowledge and the learning of new knowledge. Experimental results on four benchmark datasets demonstrate the effectiveness of our proposal and show that our method significantly outperforms other state-of-the-art incremental learning methods.
Förderung
- National Natural Science Foundation of China
- Natural Science Foundation of Jiangsu Province
Zitieren
APA
Liu, H., Yan, Z., Liu, B., Zhao, J., Zhou, Y., & Saddik, A. E. (2023). Distilled Meta-learning for Multi-Class Incremental Learning. ACM Transactions on Multimedia Computing, Communications, and Applications, 19(4), 1-16. https://doi.org/10.1145/3576045
Vancouver
Liu H, Yan Z, Liu B, Zhao J, Zhou Y, Saddik AE. Distilled Meta-learning for Multi-Class Incremental Learning. ACM Transactions on Multimedia Computing, Communications, and Applications. 2023;19(4):1-16. doi:10.1145/3576045
BibTeX
@article{Liu2023,
author = {Liu, Hao and Yan, Zhaoyu and Liu, Bing and Zhao, Jiaqi and Zhou, Yong and Saddik, Abdulmotaleb El},
title = {Distilled Meta-learning for Multi-Class Incremental Learning},
journal = {ACM Transactions on Multimedia Computing, Communications, and Applications},
year = {2023},
volume = {19},
number = {4},
pages = {1--16},
doi = {10.1145/3576045}
}
EUVIMED