Predicting Assignments of MaxSAT Variables with Graph Neural Networks
A graph neural network architecture that is trained to predict the truth value of each MaxSAT variable in an optimal solution (with a confidence score) is proposed, and significant improvements in performance are demonstrated compared to a current branch-and-bound MaxSAT solver.
J. Oliva, Peter William Nightingale, Felix Ulrich-Oltean
· 0 citations