Deep Double Descent

Study / Research

A research finding showing that in deep learning, increasing model size or data can initially hurt performance before improving it again, challenging traditional statistical notions of overfitting.

Mentioned in 1 video