Counterfactual Maximum Likelihood Estimation for Training Deep Networks

TL;DR

This paper introduces a causality-based training framework called Counterfactual Maximum Likelihood Estimation (CMLE) to reduce spurious correlations in deep networks, improving out-of-domain generalization and causal inference.

Contribution

It proposes a novel CMLE approach based on structural causal models, with algorithms for causal training using observational data, and demonstrates improved robustness in NLP and image captioning tasks.

Findings

CMLE outperforms regular MLE in out-of-domain tests

CMLE reduces spurious correlations effectively

Methods maintain comparable in-domain performance

Abstract

Although deep learning models have driven state-of-the-art performance on a wide array of tasks, they are prone to spurious correlations that should not be learned as predictive clues. To mitigate this problem, we propose a causality-based training framework to reduce the spurious correlations caused by observed confounders. We give theoretical analysis on the underlying general Structural Causal Model (SCM) and propose to perform Maximum Likelihood Estimation (MLE) on the interventional distribution instead of the observational distribution, namely Counterfactual Maximum Likelihood Estimation (CMLE). As the interventional distribution, in general, is hidden from the observational data, we then derive two different upper bounds of the expected negative log-likelihood and propose two general algorithms, Implicit CMLE and Explicit CMLE, for causal predictions of deep learning models using observational data. We conduct experiments on both simulated data and two real-world tasks: Natural Language Inference (NLI) and Image Captioning. The results show that CMLE methods outperform the regular MLE method in terms of out-of-domain generalization performance and reducing spurious correlations, while maintaining comparable performance on the regular evaluations.