Continual learning via dynamic expandable task-specific and general representations

doi:10.1016/j.knosys.2026.116851

Continual learning via dynamic expandable task-specific and general representations

2026-08-28

0

PRE

AI

DOI：10.1016/j.knosys.2026.116851

Original

Original request for help

Save

Abstract

En 中文

• A novel dual backbone structure provides general and task-specific information. • A novel CEA approach captures the shared information among all seen tasks. • A novel PBA approach promotes the new task learning. • A novel COM encourages the model to capture different semantic information.

Keywords:

Continual learning

Dynamic expansion model

Adaptive regularization optimization

Continual learning via dynamic expandable task-specific and general representations

Abstract

Journal

Organization