Gradient descent is used to reduce the loss and regularization is used to fight over-fitting.

Is there any relation between gradient descent and regularization, or both are independent of each other?


1 Answer 1


Usually, when talking about regularization for neural networks there are 3 main types: L1, L2 and dropout. All affect the gradient descent procedure.

L1 and L2 regularization is implemented in the loss function, and therefore are part of gradient descent directly by altering the derivatives of the loss function thereby altering the weight update rules of the network during gradient descent.

For L1 you add a penalty based on the $\mathcal L^1$ norm of the weight vector, while for L2 you add a penalty based on the $\mathcal L^2$ norm.

For dropout, there is no direct impact on the loss function, but you are still interfering in the gradient descent procedure indirectly by masking nodes to alter the forward and backward propagation.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .