In overparameterized linear continual learning, sequential rehearsal provably beats concurrent rehearsal for dissimilar tasks, and a hybrid of both improves accuracy and reduces forgetting on image benchmarks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Unlocking the Power of Rehearsal in Continual Learning: A Theoretical Perspective
In overparameterized linear continual learning, sequential rehearsal provably beats concurrent rehearsal for dissimilar tasks, and a hybrid of both improves accuracy and reduces forgetting on image benchmarks.