Abstract: Massive data collection holds the promise of a better understanding of complex phenomena and ultimately, of better decisions. An exciting opportunity in this regard stems from the growing availability of perturbation / intervention data (for example in genomics, advertisement, policy making, education, etc.). In order to obtain mechanistic insights from such data, a major challenge is the development of a framework that integrates observational and interventional data and allows causal transportability, i.e., predicting the effect of yet unseen interventions or transporting the effect of interventions observed in one context to another. I will propose an autoencoder framework for this problem. In particular, I will characterize the implicit bias of overparameterized autoencoders and show how this links to causal transportability and can be applied for drug repurposing in the current COVID-19 crisis.

Caroline Uhler is the Henry L. and Gracy Doherty Associate Professor in Electrical Engineering & Computer Science at the Institute for Data, Systems and Society at  Massachusetts Institute of Technology (MIT) and Assistant Professor at the Institute for Advanced Studies in Vienna. She received her PhD in Statistics from UC Berkeley in 2011 and has since received a variety of awards, including Simons Investigator Award, Sloan Research Fellowship, NSF Career Award, Sofja Kovalevskaja Prize.

Caroline Uhler is one of the world's brightest and most important internationally visible young researchers at the interface of high-dimensional statistical methods/machine learning. She works on statistical-methodological cutting-edge topics that lead to challenging optimization problems and require algorithmically sophisticated solutions. In particular, she is an expert in graphical models and causal relations, and thus in topics that are of central interest in many application problems. Her application focus at the moment is mainly biology and genetics with elementary causal biomedical problems.