Abstract: Continual learning requires models to train continuously across consecutive tasks without forgetting. Most existing methods utilize linear classifiers, which struggle to maintain a stable ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results