Reproduce “An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Uzair Akbar, Niki Kilbertus, Hao Shen, Krikamol Muandet, Bo Dai
The authors study how outcome-invariant data augmentation can reduce confounding bias by acting as a soft intervention on treatment generation. They use augmentation parameters as IV-like variables; improvement depends on alignment with spurious features. In the optical-device experiment, Gaussian-noise augmentation lowered nCER from 0.823 for ERM to 0.513 for DA+ERM; DA+IVL-CC reached 0.509, without a demonstrated significant advantage.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.