Please explain as best you can how to get the answer(with theory) and I will give you a thumbs up. Don't simply use chat GPT.

(b) Weight decay is a common regularization technique for training deep neural networks. A loss function with weight decay is given by L(w)=Lce(w)+w22, where Lce(w) is the cross-entropy loss, w is an M-dimensional vector containing all trainable weights of a deep neural network, w2 is the L2-norm of w, and >0 is a hyper-parameter controlling the degree of regularization. (i) Explain why weight decay can alleviate the overfitting problem. (5 marks) (ii) If the loss function is changed to L(w)=Lce(w)+w1, where w1=i=1Mwi is the L1-norm of w, discuss the characteristics of {wi}i=1M. When will we use the L1-norm instead of the L2-norm for weight regularization