NeurIPS 2020

Dual Instrumental Variable Regression

Meta Review

The paper studies the problem instrumental variable regression via its dual formulation. By restricting the hypothesis set to an RKHS, the authors can derive a closed-form solution to the corresponding minimax problem. The reviewers found both the theoretical results (especially the formulation in terms of the dual program) as well as the experimental results satisfying. Instrumental variables are important in causal inference and as such this paper could be of interest to the broader NeurIPS audience.