L1 and L2 regularization

I looked at it roughly the day before yesterday GAN The paper , Today I see a story about L1 A way to make the network sparse , Always wondering L1 and L2 What are you doing , And why it can be used like this .

Read a lot of explanations , I have learned a little , Here is my opinion .（ I wanted to type , It's too troublesome to find formulas and graphs , Just lazy for a while ~）