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
