Finite master programs in regularized stochastic decomposition

Simulation-based confidence bounds for two-stage stochastic programs

“…We tested the confidence bounds methods described earlier in this paper on the following test problems: APL1P (Infanger (1992) ) and PGP2 (Louveaux and Smeers (1988) ) are electric power planning problems, CEP1 (Higle and Sen (1994)) is a machine capacity expansion planning problem, STORM (Mulvey and Ruszczynski (1992)) and STORMG2, a variant of STORM, are freight scheduling problems, SCTAP1 (Ho (1980)) is a traffic assignment problem, and SNL and SNS are variants of the problem SSN (Sen, Doverspike and Cosares (1994)), a large telecomunications planning problem. Table 1 summarizes the size characteristics of each of the above problems.…”

S-model

“…The SAA estimates of Table 2 appear in Linderoth et al (2006), and these are compared with computational results from Regularized SD (Higle and Sen 1994). In addition to the four instances above, Linderoth et al (2006) also include a test instance named "gbd".…”

“…For the SD-(Average Values) column, the OBJ-LB entries correspond to the average value from solving 20 SD replications of the Regularized SD algorithm (Higle and Sen 1994). The reader might observe that there is no sample size reported for SD, because it samples until a non-parametric stopping rule (based on bootstrapping) terminates a replication of the algorithm.…”

A Probabilistic Lower Bound for Two-Stage Stochastic Programs

“…The number of universe scenarios was W = 1280. CEPl (Higle and Sen (1990) 1251) is a small machine capacity expansion planning problem (master: 9 rows and 8 columns, sub: 7 rows and 15 columns) with uncertain parameters in the right-hand side. The number of universe scenarios was W = 1000.…”