Catastrophic Inference

Definition

One of the problems that arise during the training process of an artificial neural network is catastrophic inference, in which a task being learned overwrites previous learning. As network weights are adjusted to improve performance on the