Timezone: »

 
Poster
Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minima
Guangyuan SHI · JIAXIN CHEN · Wenlong Zhang · Li-Ming Zhan · Xiao-Ming Wu

Thu Dec 09 12:30 AM -- 02:00 AM (PST) @
in Poster Session 5 »

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.

Keywords: Few Shot Learning

Author Information

Guangyuan SHI (The Hong Kong Polytechnic University)
JIAXIN CHEN (The Hong Kong Polytechnic University)
Wenlong Zhang (The Hong Kong Polytechnic University)
Li-Ming Zhan (The Hong Kong Polytechnic University)
Xiao-Ming Wu (The Hong Kong Polytechnic University)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors