Synthesizing prototypical exemplars and perturbation-based variants enables rehearsal-based continual learning with substantially fewer samples per class than standard coreset methods.
Instead of storing actual old samples, the model is jointly trained on new data and the pseudo-samples gener- ated on-demand
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Memory-efficient Continual Learning with Prototypical Exemplar Condensation
Synthesizing prototypical exemplars and perturbation-based variants enables rehearsal-based continual learning with substantially fewer samples per class than standard coreset methods.