Need help with L-2 regularization?

Question

Neural networks

There is a backpropagation neural network with a gradient descent function, the question is at what point is regularization done? I understand that regularization is applied at the stage of calculating the inverse error, but where?? when we adjust the weights?

Answer 1

Roughly speaking

newWeight = weight * (L1 + L2 * weight + grad) / batchSize