DFormer: Debiased Dual Distilled Transformer for Incremental Learning.

CoRR(2023)

引用 5|浏览32
暂无评分
摘要
In class incremental learning (CIL) setting, groups of classes are introduced to a model in each learning phase. The goal is to learn a unified model performant on all the classes observed so far. Given the recent popularity of Vision Transformers (ViTs) in conventional classification settings, an interesting question is to study their continual learning behaviour. In this work, we develop a Debiased Dual Distilled Transformer for CIL dubbed D 3 Former. The proposed model leverages a hybrid nested ViT design to ensure data efficiency and scalability to small as well as large datasets. In contrast to a recent ViT based CIL approach, our D 3 Former does not dynamically expand its architecture when new tasks are learned and remains suitable for a large number of incremental tasks. The improved CIL behaviour of D 3 Former owes to two fundamental changes to the ViT design. First, we treat the incremental learning as a long-tail classification problem where the majority samples from new classes vastly outnumber the limited exemplars available for old classes. To avoid the bias against the minority old classes, we propose to dynamically adjust logits to emphasize on retaining the representations relevant to old tasks. Second, we propose to preserve the configuration of spatial attention maps as the learning progresses across tasks. This helps in reducing catastrophic forgetting by constraining the model to retain the attention on the most discriminative regions. D 3 Former obtains favorable results on incremental versions of CIFAR-100, MNIST, SVHN, and ImageNet datasets. Code is available at https://tinyurl.com/d3former.
更多
查看译文
关键词
3 Former obtains favorable results,CIL approach,class incremental learning setting,continual learning behaviour,conventional classification settings,data efficiency,Debiased Dual Distilled Transformer,improved CIL behaviour,incremental tasks,incremental versions,learning phase,learning progresses,long-tail classification problem,minority old classes,model leverages,old tasks,recent popularity,recent ViT,unified model performant,Vision Transformers,ViT design,ViTs
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要