Learning Overlapping Representations for the Estimation of Individualized Treatment Effects
This work addresses the critical issue of personalized decision-making in fields like healthcare by improving treatment effect estimation, though it appears incremental as it builds on existing representation learning methods.
The paper tackled the problem of estimating individualized treatment effects from observational data, where selection bias and unobserved outcomes pose challenges, and demonstrated that their deep kernel regression algorithm with posterior regularization substantially outperformed state-of-the-art methods on various benchmarks.
The choice of making an intervention depends on its potential benefit or harm in comparison to alternatives. Estimating the likely outcome of alternatives from observational data is a challenging problem as all outcomes are never observed, and selection bias precludes the direct comparison of differently intervened groups. Despite their empirical success, we show that algorithms that learn domain-invariant representations of inputs (on which to make predictions) are often inappropriate, and develop generalization bounds that demonstrate the dependence on domain overlap and highlight the need for invertible latent maps. Based on these results, we develop a deep kernel regression algorithm and posterior regularization framework that substantially outperforms the state-of-the-art on a variety of benchmarks data sets.