In paper[5] Optimal Transport for Domain Adaptation, you used laplacian regularization. But I'm not sure how do we get matrix S which is a similarity matrix? Is there any tutorial related to this?

Thanks for the help.