double descent
The pattern in which a model's error falls, rises, and then falls again as its capacity is increased past the point of fitting the training data exactly.
The pattern in which a model's error falls, rises, and then falls again as its capacity is increased past the point of fitting the training data exactly.