High-Order Markov Blanket Discovery via a k-Order Relaxation of the Faithfulness Assumption

Researchers propose a k-order relaxation of the faithfulness assumption to capture higher-order dependencies such as XOR and parity relations.

They introduce kOMB, a k-order Markov blanket discovery algorithm designed to recover blankets even when standard faithfulness is violated.

The method is tested on cases with theoretical and empirical faithfulness violations, showing promise for difficult dependency structures.

This could improve Bayesian network learning, causal discovery, and feature selection in settings with noisy or parity-like relationships.