Neuroevolutionary Feature Representations for Causal Inference

TL;DR

This paper introduces a neuroevolutionary approach to learn feature representations that improve causal effect estimation by balancing outcome prediction and treatment information retention.

Contribution

It presents a novel genetic algorithm-based method for selecting neural network features that enhance heterogeneous treatment effect estimation.

Findings

Effective on synthetic data

Improves CATE estimation accuracy

Demonstrated on real-world dataset

Abstract

Within the field of causal inference, we consider the problem of estimating heterogeneous treatment effects from data. We propose and validate a novel approach for learning feature representations to aid the estimation of the conditional average treatment effect or CATE. Our method focuses on an intermediate layer in a neural network trained to predict the outcome from the features. In contrast to previous approaches that encourage the distribution of representations to be treatment-invariant, we leverage a genetic algorithm that optimizes over representations useful for predicting the outcome to select those less useful for predicting the treatment. This allows us to retain information within the features useful for predicting outcome even if that information may be related to treatment assignment. We validate our method on synthetic examples and illustrate its use on a real life dataset.