Abstract: We present a bag of tricks framework for few-shot class-incremental learning (FSCIL), which is a challenging form of continual learning that involves continuous adaptation to new tasks with limited samples. FSCIL requires both stability and adaptability, i.e., preserving proficiency in previously learned tasks while learning new ones. Our proposed bag of tricks brings together six key and highly influential techniques that improve stability, adaptability, and overall performance under a unified framework for FSCIL. We organize these tricks into three categories: stability tricks, adaptability tricks, and training tricks. Stability tricks aim to mitigate the forgetting of previously learned classes by enhancing the separation between the embeddings of learned classes and minimizing interference when learning new ones. On the other hand, adaptability tricks focus on the effective learning of new classes. Finally, training tricks improve the overall performance without compromising stability or adaptability. We perform extensive experiments on three benchmark datasets, CIFAR-100, CUB-200, and miniIMageNet, to evaluate the impact of our proposed framework. Our detailed analysis shows that our approach substantially improves both stability and adaptability, establishing a new state-of-the-art by outperforming prior works in the area. We believe our method provides a go-to solution and establishes a robust baseline for future research in this area.
Submission Length: Regular submission (no more than 12 pages of main content)
Changes Since Last Submission: As per the suggestion, we have now included additional experiments that leverage pre-trained large models as our backbone. More specifically, we have expanded Table 3 with the pre-trained ViT-B/32 and ViT-B/14. The results show a similar trend in performance as reported previously with other encoders.
We believe these additional results strengthen our findings.
Assigned Action Editor: ~Lijun_Zhang1
Submission Number: 2829
Loading