Learning-Augmented Algorithms for Online Steiner Tree

This paper considers the recently popular beyond-worst-case algorithm analysis model which integrates machine-learned predictions with online design. We consider Steiner tree problem in this for both directed and undirected graphs. is known to have strong lower bounds setting any algorithm’s worst-case guarantee far from desirable. algorithms that predict terminal arrives online. The may be incorrect algorithms’ performance parameterized by number of incorrectly predicted terminals. These guarantees ensure break through good competitive ratio gracefully degrades as prediction error grows. then observe theory predictive what will occur empirically. show on graphs where terminals are drawn a distribution, new even modestly correct predictions.