Catastrophic forgetting
Catastrophic forgetting: Loss of performance on previous knowledge when a model is updated on new data or tasks. It is central to continual learning. Replay, regularisation, separate modules and retrospective evaluations help preserve acquired capabilities. This perspective connects the definition with practical behaviour in artificial intelligence systems.
In italiano: Catastrophic ForgettingLegge III · Come una rete impara dai datiIt is central to continual learning. Replay, regularisation, separate modules and retrospective evaluations help preserve acquired capabilities.