A soft nearest-neighbor framework for continual semi-supervised learning

Zhiqi Kang (1) , Enrico Fini (2) , Moin Nabi (3) , Elisa Ricci (2, 4) , Karteek Alahari (1)

1 Thoth - Apprentissage de modèles à partir de données massives
2 UNITN - Università degli Studi di Trento = University of Trento
3 SAP AI - SAP AI Research
4 FBK - Fondazione Bruno Kessler [Trento, Italy]

Despite significant advances, the performance of state-of-the-art continual learning approaches hinges on the unrealistic scenario of fully labeled data. In this paper, we tackle this challenge and propose an approach for continual semi-supervised learning---a setting where not all the data samples are labeled. An underlying issue in this scenario is the model forgetting representations of unlabeled data and overfitting the labeled ones. We leverage the power of nearest-neighbor classifiers to non-linearly partition the feature space and learn a strong representation for the current task, as well as distill relevant information from previous tasks. We perform a thorough experimental evaluation and show that our method outperforms all the existing approaches by large margins, setting a strong state of the art on the continual semi-supervised learning paradigm. For example, on CIFAR100 we surpass several others even when using at least 30 times less supervision (0.8% vs. 25% of annotations). The code is publicly available on https://github.com/kangzhiq/NNCSL

Keywords: Continual learning; Semi-supervised learning; Image classification

Domains: Computer Vision and Pattern Recognition [cs.CV]

Zhiqi Kang, Enrico Fini, Moin Nabi, Elisa Ricci, Karteek Alahari. A soft nearest-neighbor framework for continual semi-supervised learning. 2022. ⟨hal-03893056v1⟩