Generalized Counting for Lifted Variable Elimination

摘要: Lifted probabilistic inference methods exploit symmetries in the structure of models to perform more efficiently. In lifted variable elimination, symmetry among a group interchangeable random variables is captured by counting formulas, and exploited operations that handle such formulas. this paper, we generalize formulas present set operators introduce eliminate these from model. This generalization expands range problems can be solved way. Our work closely related recently introduced method joint conversion. Due its fine grained formulation, however, our approach provide efficient solutions than