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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Although the concept of catastrophic forgetting is straightforward, there isa lack of study on its causes. In this paper, we systematically explore andreveal three causes for catastrophic forgetting in Class IncrementalLearning(CIL). From the perspective of representation learning,(i) intra-phaseforgetting happens when the learner fails to correctly align the same-phasedata as training proceeds and (ii) inter-phase confusion happens when thelearner confuses the current-phase data with the previous-phase. From thetask-specific point of view, the CIL model suffers from the problem of (iii)classifier deviation. After investigating existing strategies, we observe thatthere is a lack of study on how to prevent the inter-phase confusion. Toinitiate the research on this specific issue, we propose a simple yet effectiveframework, Contrastive Class Concentration for CIL (C4IL). Our frameworkleverages the class concentration effect of contrastive learning, yielding arepresentation distribution with better intra-class compactibility andinter-class separability. Empirically, we observe that C4IL significantlylowers the probability of inter-phase confusion and as a result improves theperformance on multiple CIL settings of multiple datasets.

Revisiting Catastrophic Forgetting in Class Incremental Learning