Two-stage stochastic matching and spanning tree problems: Polynomial instances and approximation

This article deals with the two-stage stochastic model, which aims at explicitly taking into account uncertainty in optimization problems, that Kong and Schaefer have recently studied for the maximum weight matching problem (2006). They have proved that the problem is NP-hard, and they have provided a factor ½ approximation algorithm. We further study this problem and strengthen the hardness results, slightly improve the approximation ratio and exhibit some polynomial cases. We similarly tackle the maximum weight spanning tree problem in the two-stage setting. Finally, we make numerical experiments on randomly generated instances to compare the quality of several interesting heuristics.