Spectral Representation for Causal Estimation with Hidden Confounders
It addresses a key challenge in causal inference for researchers and practitioners, but appears incremental as it builds on prior saddle-point methods.
The paper tackles causal effect estimation with hidden confounders in instrumental variable regression and proxy causal learning settings, using a spectral representation and saddle-point optimization, and shows that it outperforms existing methods on benchmarks.
We address the problem of causal effect estimation where hidden confounders are present, with a focus on two settings: instrumental variable regression with additional observed confounders, and proxy causal learning. Our approach uses a singular value decomposition of a conditional expectation operator, followed by a saddle-point optimization problem, which, in the context of IV regression, can be thought of as a neural net generalization of the seminal approach due to Darolles et al. [2011]. Saddle-point formulations have gathered considerable attention recently, as they can avoid double sampling bias and are amenable to modern function approximation methods. We provide experimental validation in various settings, and show that our approach outperforms existing methods on common benchmarks.