24
1

Learning-augmented Maximum Independent Set

Abstract

We study the Maximum Independent Set (MIS) problem on general graphs within the framework of learning-augmented algorithms. The MIS problem is known to be NP-hard and is also NP-hard to approximate to within a factor of n1δn^{1-\delta} for any δ>0\delta>0. We show that we can break this barrier in the presence of an oracle obtained through predictions from a machine learning model that answers vertex membership queries for a fixed MIS with probability 1/2+ε1/2+\varepsilon. In the first setting we consider, the oracle can be queried once per vertex to know if a vertex belongs to a fixed MIS, and the oracle returns the correct answer with probability 1/2+ε1/2 + \varepsilon. Under this setting, we show an algorithm that obtains an O~(Δ/ε)\tilde{O}(\sqrt{\Delta}/\varepsilon)-approximation in O(m)O(m) time where Δ\Delta is the maximum degree of the graph. In the second setting, we allow multiple queries to the oracle for a vertex, each of which is correct with probability 1/2+ε1/2 + \varepsilon. For this setting, we show an O(1)O(1)-approximation algorithm using O(n/ε2)O(n/\varepsilon^2) total queries and O~(m)\tilde{O}(m) runtime.

View on arXiv
Comments on this paper