Class-Incremental Learning (CIL) aims to solve the neural networks' catastrophic forgetting problem, which refers to the fact that once the network updates on a new task, its performance on previously-learned tasks drops dramatically. Most successful CIL methods incrementally train a feature extractor with the aid of stored exemplars, or estimate the feature distribution with the stored prototypes. However, the stored exemplars would violate the data privacy concerns, while the stored prototypes might not reasonably be consistent with a proper feature distribution, hindering the exploration of real-world CIL applications. In this paper, we propose a method of \textit{e}mbedding distillation and \textit{Ta}sk-oriented \textit{g}eneration (\textit{eTag}) for CIL, which requires neither the exemplar nor the prototype. Instead, eTag achieves a data-free manner to train the neural networks incrementally. To prevent the feature extractor from forgetting, eTag distills the embeddings of the network's intermediate blocks. Additionally, eTag enables a generative network to produce suitable features, fitting the needs of the top incremental classifier. Experimental results confirmed that our proposed eTag considerably outperforms the state-of-the-art methods on CIFAR-100 and ImageNet-sub\footnote{Our code is available in the Supplementary Materials.
翻译:类增量学习(Class-Incremental Learning,CIL)旨在解决神经网络“灾难性遗忘”问题,即一旦网络更新了一个新任务,它在先前学习的任务上的性能会大幅度下降。大多数成功的CIL方法会利用存储的范例递增地训练特征提取器,或者利用存储的原型估计特征分布。然而,存储的范例会侵犯数据隐私,而存储的原型可能不合理地与适当的特征分布一致,从而阻碍了对实际CIL应用的探索。本文提出了一种名为嵌入式提取和任务导向生成(eTag)的CIL方法,它既不需要范例,也不需要原型。相反,eTag以一种无需数据的方式递增地训练神经网络。为防止特征提取器遗忘,eTag对网络的中间块的嵌入进行了提取。此外,eTag还使生成网络能够产生适合顶部增量分类器需求的特征。实验结果证实,我们提出的eTag在CIFAR-100和ImageNet-sub上显著优于现有的最先进方法。我们的代码可在补充材料中获取。