Concetti Chiave
The authors propose a new encoding technique that reduces the exponential dependency of the polynomial degree on the counting quantifier parameters in the time complexity bound for computing weighted first-order model counting over the two-variable fragment with counting quantifiers.
Sintesi
The paper studies the time complexity of weighted first-order model counting (WFOMC) over the logical language with two variables and counting quantifiers, known as the C2 fragment.
The key contributions are:
The authors derive an upper bound on the time complexity of computing WFOMC over C2 using existing techniques. They show that the polynomial degree of the bound depends exponentially on the parameters of the counting quantifiers appearing in the input formula.
To address this issue, the authors propose a new encoding technique that reduces the exponential dependency to a quadratic one, thereby obtaining a tighter upper bound on the time complexity.
The new encoding leverages the concept of "canonical models" to count only a representative set of models, while properly weighting them to recover the correct WFOMC. This approach significantly improves the previous upper bound.
The authors also provide experimental results demonstrating the practical benefits of their new encoding compared to the existing techniques.