Exact stochastic constraint optimisation with applications in network analysis

We present an extensive study of methods for exactly solving stochastic constraint (optimisation) problems (SCPs) in network analysis. These are prevalent science, governance and industry. The first method we is generic decomposes constraints into a multitude smaller local that solved using programming (CP) or mixed-integer (MIP) solver. However, many SCPs formulated on probability distributions with monotonic property, meaning adding positive decision to partial solution the problem cannot cause decrease quality. second specifically designed global (SCMDs) CP. Both use knowledge compilation obtain diagram encoding relevant distributions, where focus ordered binary diagrams (OBDDs). discuss theoretical advantages disadvantages these evaluate them experimentally. observed approaches SCMDs outperform decomposition from CP, perform complementarily MIP-based approaches, while scaling much more favourably instance size. have alternative design choices, as both solvers used single pipeline. To identify which configurations work best, apply by optimisation. Specifically, show how automated algorithm configurator can be find optimised our After configuration, SCMD pipeline outperforms its closest competitor (a pipeline) all test sets considered up two orders magnitude terms PAR10 scores.