ConceptReplay Mixing / Replay
Replay Mixing
A regularization strategy in continual learning that retains a portion of old data and assigns it a higher sampling weight when fine-tuning on a new task, thereby mitigating catastrophic forgetting.
Timeline (last 90 days)
Oct 17
Replay mixing is a regularization strategy in continual learning: when fine-tuning for a new task or domain, a portion of old data is deliberately retained and given a higher sampling weight to prevent the model from forgetting old capabilities.
Unverified50%