English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Continual learning aims to rapidly and continually learn the current taskfrom a sequence of tasks.  Compared to other kinds of methods, the methods based on experience replayhave shown great advantages to overcome catastrophic forgetting. One commonlimitation of this method is the data imbalance between the previous andcurrent tasks, which would further aggravate forgetting. Moreover, how toeffectively address the stability-plasticity dilemma in this setting is also anurgent problem to be solved. In this paper, we overcome these challenges byproposing a novel framework called Meta-learning update via Multi-scaleKnowledge Distillation and Data Augmentation (MMKDDA). Specifically, we applymultiscale knowledge distillation to grasp the evolution of long-range andshort-range spatial relationships at different feature levels to alleviate theproblem of data imbalance. Besides, our method mixes the samples from theepisodic memory and current task in the online continual training procedure,thus alleviating the side influence due to the change of probabilitydistribution. Moreover, we optimize our model via the meta-learning updateresorting to the number of tasks seen previously, which is helpful to keep abetter balance between stability and plasticity. Finally, our experimentalevaluation on four benchmark datasets shows the effectiveness of the proposedMMKDDA framework against other popular baselines, and ablation studies are alsoconducted to further analyze the role of each component in our framework.

Online Continual Learning via the Meta-learning Update with Multi-scale Knowledge Distillation and Data Augmentation