On the Conditional Independence Implication Problem: A Lattice-Theoretic Approach
This work addresses a foundational problem in probability theory and machine learning, offering incremental improvements in computational methods for CI inference.
The paper tackles the conditional independence implication problem for discrete probability measures by introducing a lattice-theoretic framework, resulting in a sound and complete inference system for certain CI statements and polynomial-time heuristics for falsification.
A lattice-theoretic framework is introduced that permits the study of the conditional independence (CI) implication problem relative to the class of discrete probability measures. Semi-lattices are associated with CI statements and a finite, sound and complete inference system relative to semi-lattice inclusions is presented. This system is shown to be (1) sound and complete for saturated CI statements, (2) complete for general CI statements, and (3) sound and complete for stable CI statements. These results yield a criterion that can be used to falsify instances of the implication problem and several heuristics are derived that approximate this "lattice-exclusion" criterion in polynomial time. Finally, we provide experimental results that relate our work to results obtained from other existing inference algorithms.