Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper investigates the utility of Markov boundaries for tabular prediction on the SCM3K benchmark, comparing oracle boundaries to those recovered via causal discovery. They find that while using the oracle Markov boundary often improves prediction, especially in high-dimensional and sparse feature spaces, recovering the boundary with existing causal discovery methods and then training on it does not consistently outperform using the full feature set. The authors attribute this discrepancy to the mismatch between structural recovery and prediction objectives, asymmetric costs of false positives/negatives, and the existence of multiple effective feature sets beyond the exact Markov boundary.
Causal discovery's focus on structural accuracy, rather than predictive performance, renders Markov boundaries surprisingly ineffective for improving tabular prediction in practice.
Under standard graphical assumptions, the Markov boundary of a target variable is the smallest set of features that renders every other feature redundant. Once the boundary is observed, the target is conditionally independent of the rest of the table. This is a tempting object for tabular prediction, since it names exactly the columns a model should need. Yet modern regressors are still trained on the full feature set. We ask whether the Markov boundary is genuinely useful for prediction on SCM3K, a 3,450-task synthetic SCM benchmark with feature counts from 40 to 1000 and six SCM families, evaluated with six regressors. The answer is more nuanced than the theory suggests. Restricting a regressor to the oracle boundary often improves prediction substantially, and the improvement grows as the feature space becomes larger and sparser. But the natural pipeline of recovering the boundary with causal discovery and training on the recovered mask does not deliver. Existing estimators exhaust the compute budget before reaching the regime where the boundary helps most, and even where they run they rarely beat the full feature set. We trace this to three causes. Discovery optimizes structural recovery rather than prediction. False negatives and false positives carry sharply asymmetric predictive cost. The exact boundary is only one of many feature sets that beat all features. We then develop what these facts imply for prediction-aligned feature selection and for tabular models that learn to use causal structure.