Abstract: Rehearsal methods based on knowledge distillation (KD) have been widely used in continual learning (CL). However, given memory constraints, few exemplars contain limited variations of ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback