SFA reduces forgetting in continual learning by averaging the currently training model with a checkpoint from previous tasks during fine-tuning, matching data-buffer rehearsal without storing past data.
But how severe is this forgetting? We quantify this by comparing a model that was trained on and has then forgotten Simpl to a model that has never seen Simpl
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
-
Soup to go: mitigating forgetting during continual learning with model averaging
SFA reduces forgetting in continual learning by averaging the currently training model with a checkpoint from previous tasks during fine-tuning, matching data-buffer rehearsal without storing past data.