Happy new year and welcome back to NYOSP meetups. To begin the year we have Sean Taylor (author of the much talked about blog post on the Statistics Software Signal) discussing ranking algorithms.
About his talk:
Ranking elements from a set of objects is a common statistical problem with applications ranging from web search and recommendation systems to modeling individual preferences or social hierarchies. We all know--and probably use--approaches such as PageRank or logistic regression to create orderings of objects. In this talk, I will describe the formal learning problem involved in inferring rankings which minimize different loss functions. We'll then take a helicopter tour through a few common ranking techniques and apply them to example ranking problems from sports, social networks, and information retrieval. I will also spend some time discussing regularization and active learning in these contexts, as well as when ranking models provide a bad approximation of reality.
Sean J. Taylor is a 5th year PhD candidate at NYU's Stern School of Business. His work combines randomized experiments and Bayesian statistics to learn about social influence processes in the online world. In addition to his work at NYU, he has conducted research at The Wharton School, the US Federal Reserve Board, and most recently with Facebook's Data Science team. In his spare time, he analyzes NFL data to win his fantasy football league and find reasons to be optimistic about the Eagles.
Even though he is an Eagles fan we'll still allow him to have pizza (starting at 6:15) and join us at the bar after the talk which will start around 7.