Abstract: One of the biggest promises of machine learning is the automation of decision making in a multitude of application domains. A core problem that arises in most data-driven personalized decision scenarios is the estimation of heterogeneous treatment effects: what is the effect of an intervention on an outcome of interest as a function of a set of observable characteristics of the treated sample? For instance, this problem arises in personalized pricing, where the goal is to estimate the effect of a price discount on the demand as a function of characteristics of the consumer. Similarly it arises in medical trials where the goal is to estimate the effect of a drug treatment on the clinical response of a patient as a function of patient characteristics. In many such settings we have an abundance of observational data, where the intervention was chosen via some unknown policy and the ability to run control A/B tests is limited.
We will present recent research advances in the area of machine learning based estimation of heterogeneous treatment effects. These novel methods offer large flexibility in modeling the effect heterogeneity (via techniques such as random forests, boosting, lasso and neural nets), while at the same time leverage techniques from causal inference and econometrics to preserve the causal interpretation of the learned model and many times also offer statistical validity via the construction of valid confidence intervals. We will also present and demo the Microsoft EconML library, an open source package developed by the ALICE project of Microsoft Research, New England, which implements several recent estimation algorithms in a common python API.
Bio: Vasilis Syrgkanis is a Researcher at Microsoft Research, New England. He received his Ph.D. in Computer Science from Cornell University in 2014, under the supervision of Prof. Eva Tardos and subsequently spend two years in Microsoft Research, New York as a postdoctoral researcher in the Machine Learning and Algorithmic Economics groups. His research addresses problems at the intersection of theoretical computer science, machine learning and economics. His work received best paper awards at the 2015 ACM Conference on Economics and Computation (EC'15) and at the 2015 Annual Conference on Neural Information Processing Systems (NIPS'15) and was the recipient of the Simons Fellowship for graduate students in theoretical computer science 2012-2014.