weight decay

A regularization technique used in training neural networks to prevent overfitting by adding a penalty to the loss function based on the magnitude of the weights. This encourages simpler models that generalize better on unseen data.

12 papers