eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation
Proceedings of the AAAI Conference on Artificial Intelligence · 24.03.2024 · 38 (11) · 12591-12599
DOI: 10.1609/aaai.v38i11.29153 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 24.03.2024. Verfasst von Libo Huang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 28-mal zitiert (Crossref).
Abstract (Originalsprache)
Class incremental learning (CIL) aims to solve the notorious forgetting problem, which refers to the fact that once the network is updated on a new task, its performance on previously-learned tasks degenerates catastrophically. Most successful CIL methods store exemplars (samples of learned tasks) to train a feature extractor incrementally, or store prototypes (features of learned tasks) to estimate the incremental feature distribution. However, the stored exemplars would violate the data privacy concerns, while the fixed prototypes might not reasonably be consistent with the incremental feature distribution, hindering the exploration of real-world CIL applications. In this paper, we propose a data-free CIL method with embedding distillation and Task-oriented generation (eTag), which requires neither exemplar nor prototype. Embedding distillation prevents the feature extractor from forgetting by distilling the outputs from the networks' intermediate blocks. Task-oriented generation enables a lightweight generator to produce dynamic features, fitting the needs of the top incremental classifier. Experimental results confirm that the proposed eTag considerably outperforms state-of-the-art methods on several benchmark datasets.
Zitieren
APA
Huang, L., Zeng, Y., Yang, C., An, Z., Diao, B., & Xu, Y. (2024). eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation. Proceedings of the AAAI Conference on Artificial Intelligence, 38(11), 12591-12599. https://doi.org/10.1609/aaai.v38i11.29153
Vancouver
Huang L, Zeng Y, Yang C, An Z, Diao B, Xu Y. eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation. Proceedings of the AAAI Conference on Artificial Intelligence. 2024;38(11):12591-12599. doi:10.1609/aaai.v38i11.29153
BibTeX
@article{Huang2024,
author = {Huang, Libo and Zeng, Yan and Yang, Chuanguang and An, Zhulin and Diao, Boyu and Xu, Yongjun},
title = {eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2024},
volume = {38},
number = {11},
pages = {12591--12599},
doi = {10.1609/aaai.v38i11.29153}
}
EUVIMED