Abstract
Classical Recommender Systems provide serving schemes that display items to users to optimize some user satisfaction metrics. But what happens when such users are connected to each other as in social media platforms like Facebook, LinkedIn and Twitter? Content in such a network is produced by nodes (users) and flows through edges in the graph. This gives rise to challenging and brand new issues when dealing with classical problems like recommending content to users. I will talk about such issues, based on my experience working with feed recommendation problems at LinkedIn.