PT Unknown
AU Kai Wang
   Xialei Liu
   Andrew Bagdanov
   Luis Herranz
   Shangling Jui
   Joost Van de Weijer
TI Incremental Meta-Learning via Episodic Replay Distillation for Few-Shot Image Recognition
BT CVPR 2022 Workshop on Continual Learning (CLVision, 3rd Edition)
PY 2022
BP 3728
EP 3738
DI 10.1109/CVPRW56347.2022.00417
DE Training; Computer vision; Image recognition; Upper bound; Conferences; Pattern recognition; Task analysis
AB In this paper we consider the problem of incremental meta-learning in which classes are presented incrementally in discrete tasks. We propose Episodic Replay Distillation (ERD), that mixes classes from the current task with exemplars from previous tasks when sampling episodes for meta-learning. To allow the training to benefit from a large as possible variety of classes, which leads to more gener-alizable feature representations, we propose the cross-task meta loss. Furthermore, we propose episodic replay distillation that also exploits exemplars for improved knowledge distillation. Experiments on four datasets demonstrate that ERD surpasses the state-of-the-art. In particular, on the more challenging one-shot, long task sequence scenarios, we reduce the gap between Incremental Meta-Learning andthe joint-training upper bound from 3.5% / 10.1% / 13.4% / 11.7% with the current state-of-the-art to 2.6% / 2.9% / 5.0% / 0.2% with our method on Tiered-ImageNet / Mini-ImageNet / CIFAR100 / CUB, respectively.
ER