TY - GEN
T1 - Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minima
AU - Shi, Guangyuan
AU - Chen, Jiaxin
AU - Zhang, Wenlong
AU - Zhan, Li Ming
AU - Wu, Xiao Ming
N1 - Funding Information:
We would like to thank the anonymous reviewers for their insightful and helpful comments. This research was supported by the grant of DaSAIL project P0030935 funded by PolyU/UGC.
Publisher Copyright:
© 2021 Neural information processing systems foundation. All rights reserved.
PY - 2021/11
Y1 - 2021/11
N2 - This paper considers incremental few-shot learning, which requires a model to continually recognize new categories with only a few examples provided. Our study shows that existing methods severely suffer from catastrophic forgetting, a well-known problem in incremental learning, which is aggravated due to data scarcity and imbalance in the few-shot setting. Our analysis further suggests that to prevent catastrophic forgetting, actions need to be taken in the primitive stage – the training of base classes instead of later few-shot learning sessions. Therefore, we propose to search for flat local minima of the base training objective function and then fine-tune the model parameters within the flat region on new tasks. In this way, the model can efficiently learn new classes while preserving the old ones. Comprehensive experimental results demonstrate that our approach outperforms all prior state-of-the-art methods and is very close to the approximate upper bound. The source code is available at https://github.com/moukamisama/F2M.
AB - This paper considers incremental few-shot learning, which requires a model to continually recognize new categories with only a few examples provided. Our study shows that existing methods severely suffer from catastrophic forgetting, a well-known problem in incremental learning, which is aggravated due to data scarcity and imbalance in the few-shot setting. Our analysis further suggests that to prevent catastrophic forgetting, actions need to be taken in the primitive stage – the training of base classes instead of later few-shot learning sessions. Therefore, we propose to search for flat local minima of the base training objective function and then fine-tune the model parameters within the flat region on new tasks. In this way, the model can efficiently learn new classes while preserving the old ones. Comprehensive experimental results demonstrate that our approach outperforms all prior state-of-the-art methods and is very close to the approximate upper bound. The source code is available at https://github.com/moukamisama/F2M.
UR - http://www.scopus.com/inward/record.url?scp=85131890834&partnerID=8YFLogxK
M3 - Conference article published in proceeding or book
AN - SCOPUS:85131890834
T3 - Advances in Neural Information Processing Systems
SP - 6747
EP - 6761
BT - Advances in Neural Information Processing Systems 34 - 35th Conference on Neural Information Processing Systems, NeurIPS 2021
A2 - Ranzato, Marc'Aurelio
A2 - Beygelzimer, Alina
A2 - Dauphin, Yann
A2 - Liang, Percy S.
A2 - Wortman Vaughan, Jenn
PB - Neural information processing systems foundation
T2 - 35th Conference on Neural Information Processing Systems, NeurIPS 2021
Y2 - 6 December 2021 through 14 December 2021
ER -