What is catastrophic forgetting in neural networks?

Catastrophic forgetting occurs when models lose old knowledge while learning new data.

In continual learning, new training overwrites old weights. Solutions include replay buffers, regularization, and progressive networks.