What is ‘catastrophic forgetting’ in neural networks and how is it addressed in continual learning?

Question

AI Fundamentals — Hard

What is ‘catastrophic forgetting’ in neural networks and how is it addressed in continual learning?

Accepted Answer

Catastrophic forgetting refers to neural networks losing previously learned information when trained on new tasks. This is addressed in continual learning through techniques like elastic weight consolidation, replay buffers, or progressive networks.