A column-and-constraint generation algorithm for two-stage stochastic programming problems

Abstract This paper presents a column-and-constraint generation algorithm for two-stage stochastic programming problems. A distinctive feature of the is that it does not assume fixed recourse and as consequence values dimensions matrix can be uncertain. The proposed contains multi-cut (partial) Benders decomposition deterministic equivalent model special cases used to trade-off computational speed memory requirements. outperforms in time requirements maintenance location routing problem. In addition, instances with large number scenarios, both Furthermore, we present an adaptive relative tolerance which solution master problem bottleneck slave problems solved relatively efficiently. early iterations converges zero final iteration(s) algorithm. combination this decreases our even further.