Automated Hyperparameter Selection for the PC Algorithm

TL;DR

AutoPC is a novel method that automatically optimizes the significance level in the PC causal discovery algorithm, improving stability and performance without traditional cross-validation.

Contribution

We introduce AutoPC, a fast, automated hyperparameter tuning procedure for the PC algorithm that enhances causal graph accuracy by maximizing stability between repeated runs.

Findings

AutoPC outperforms existing methods across multiple metrics.

It improves the stability of causal graph recovery.

AutoPC requires no traditional cross-validation.

Abstract

The PC algorithm infers causal relations using conditional independence tests that require a pre-specified Type I $\alpha$ level. PC is however unsupervised, so we cannot tune $\alpha$ using traditional cross-validation. We therefore propose AutoPC, a fast procedure that optimizes $\alpha$ directly for a user chosen metric. We in particular force PC to double check its output by executing a second run on the recovered graph. We choose the final output as the one which maximizes stability between the two runs. AutoPC consistently outperforms the state of the art across multiple metrics.