PhD student Eli Sherman will present his research at the 2018 Conference on Neural Information Processing Systems (NeurIPS), held December 3-8 in Montreal, Canada.
NeurIPS is the largest academic conference on artificial intelligence (AI) and brings together researchers in machine learning, computational neuroscience, and statistics.
The conference receives thousands of paper submissions each year. This year, Sherman was selected to give a spotlight presentation on his paper, “Identification and Estimation Of Causal Effects from Dependent Data.”
Sherman works in the Hopkins Causal Decision Making Group, under the supervision of John C. Malone Assistant Professor Ilya Shpitser. He is working on the development of graphical model-based methods for making causal inferences, as well as the application of machine learning to causal inference with a special focus on health care and missing data.
From the abstract:
“The assumption that data samples are independent and identically distributed (iid) is standard in many areas of statistics and machine learning. Nevertheless, in some settings, such as social networks, infectious disease modeling, and reasoning with spatial and temporal data, this assumption is false. An extensive literature exists on making causal inferences under the iid assumption, even when unobserved confounding bias may be present. But, causal inference in non-iid contexts is challenging due to the presence of both unobserved confounding and data dependence. In this paper we develop a general theory describing when causal inferences are possible in such scenarios. We use segregated graphs, a generalization of latent projection mixed graphs, to represent causal models of this type and provide a complete algorithm for nonparametric identification in these models. We then demonstrate how statistical inferences may be performed on causal parameters identified by this algorithm. In particular, we consider cases where only a single sample is available for parts of the model due to full interference, i.e., all units are pathwise dependent and neighbors’ treatments affect each others outcomes. We apply these techniques to a synthetic data set which considers the adoption of fake news articles given the social network structure, articles read by each person, and baseline demographics and socioeconomic covariates.”