Description
In this talk Maurits will introduce the problem of personalization and discuss some its peculiarities.
Over ten years Maurits has been exploring different statistical methods to efficiently learn “what to do to whom”. His work has been applied various fields such as marketing (who should we target with our retention campaigns?) and healthcare (which type of treatment has the best expected outcome for the current patient?). In this talk Maurits will introduce the problem of personalization and discuss some its peculiarities. Maurits will focus on two contemporary challenges: First, how should we — if we are lucky enough to be able to — setup experiments to learn “what to do to whom” through sequential experimentation? Second, how can we learn, based on already existing data, what we should do to whom in the future? This talk will cover some of the intuitions behind effective sequential learning and offline policy evaluation, and will highlight a number of tools and software packages developed by member of Maurits’ lab that enable you to get started right away.
Lab website: www.nth-iteration.com