Resumen

GUTIERREZ-DE-LA-PAZ, Omar et al. Policy Gradient MaxSAT Solver. Comp. y Sist. [online]. 2024, vol.28, n.2, pp.353-366.  Epub 31-Oct-2024. ISSN 2007-9737.  https://doi.org/10.13053/cys-28-2-4723.

This paper presents a comparative study of various elements and strategies that can be incorporated into an autoregressive model to address the MaxSAT problem. Building upon a sequential architecture as our foundation, we optimize the model’s parameters by maximizing the expected number of satisfied clauses. This optimization enables the model, given a SAT formula, to predict a distribution over potential solutions using the policy gradient method. Our controlled experiments pinpoint elements that guide the optimization process towards superior resultsfn.

Palabras llave : MaxSAT problem; policy gradient; NP-hard.

        · texto en Inglés     · Inglés ( pdf )