I2CANSAY:Inter-Class Analogical Augmentation and Intra-Class Significance Analysis for Non-Exemplar Online Task-Free Continual Learning

TL;DR

A novel framework called I$^{2}$CANSAY that gets rid of the dependence on memory buffers and efficiently learns the knowledge of new data from one-shot samples and outperforms the prior state-of-the-art by a large margin.

Abstract

Online task-free continual learning (OTFCL) is a more challenging variant of continual learning which emphasizes the gradual shift of task boundaries and learns in an online mode. Existing methods rely on a memory buffer composed of old samples to prevent forgetting. However,the use of memory buffers not only raises privacy concerns but also hinders the efficient learning of new samples. To address this problem, we propose a novel framework called I2CANSAY that gets rid of the dependence on memory buffers and efficiently learns the knowledge of new data from one-shot samples. Concretely, our framework comprises two main modules. Firstly, the Inter-Class Analogical Augmentation (ICAN) module generates diverse pseudo-features for old classes based on the inter-class analogy of feature distributions for different new classes, serving as a substitute for the memory buffer. Secondly, the Intra-Class Significance Analysis (ISAY) module analyzes the significance of attributes for each class via its distribution standard deviation, and generates the importance vector as a correction bias for the linear classifier, thereby enhancing the capability of learning from new samples. We run our experiments on four popular image classification datasets: CoRe50, CIFAR-10, CIFAR-100, and CUB-200, our approach outperforms the prior state-of-the-art by a large margin.

Figures (7)

• Figure 1: The motivation of (a) inter-class analogical augmentation and (b) intra-class significance analysis module.
• Figure 2: The framework of our model. Our model is composed of two modules. The ICAN module generates diverse pseudo-features via prototype $P$ and STDs $R$ and trains the classifier's weights $w_g$ by incorporating them along with real features. The ISAY module analyzes the intra-class significance attributes for each class via STDs $R$ and generates the importance vector $\tau_x$ as the correction bias for the classifier.
• Figure 3: The illustration of analogical pseudo features generation. ICAN module transfers the feature distribution of new data proportionally to the old classes.
• Figure 4: The distributions of features in dimensions a and b for the classes 'fox' and 'lamp', using the dino$\_$ViT8 feature extractor.
• Figure 5: The OTFCL session accuracy in Split CIFAR-10 (step=2).