Exemplar-free Continual Representation Learning via Learnable Drift Compensation

The code will be available after ECCV

This is the official repository for the paper:

Exemplar-free Continual Representation Learning via Learnable Drift Compensation
Alex Gomez-Villa, Dipam Goswami, Kai Wang, Andrew D. Bagdanov, Bartlomiej Twardowski, Joost van de Weijer
ECCV 2024

Abstract: Exemplar-free class-incremental learning using a backbone trained from scratch and starting from a small first task presents a significant challenge for continual representation learning. Prototype-based approaches, when continually updated, face the critical issue of semantic drift due to which the old class prototypes drift to different positions in the new feature space. Through an analysis of prototype-based continual learning, we show that forgetting is not due to diminished discriminative power of the feature extractor, and can potentially be corrected by drift compensation. To address this, we propose Learnable Drift Compensation (LDC), which can effectively mitigate drift in any moving backbone, whether supervised or unsupervised. LDC is fast and straightforward to integrate on top of existing continual learning approaches. Furthermore, we showcase how LDC can be applied in combination with self-supervised CL methods, resulting in the first exemplar-free semi-supervised continual learning approach. We achieve state-of-the-art performance in both supervised and semi-supervised settings across multiple datasets.

Citation

If you like our work, please cite our paper:

@inproceedings{gomezvilla2024, 
    title={Exemplar-free Continual Representation Learning via Learnable Drift Compensation}, 
    author={Gomez-Villa, Alex and Goswami, Dipam and Wang, Kai and Bagdanov Andrew and 
            Twardowski, Bartlomiej  and van de Weijer, Joost},
    booktitle={European Conference on Computer Vision}, 
    year={2024}
}