- What is POSTS?
- How to get POSTS?
- Suggested reporting rules for POSTS problems?
- Problem Descriptions
- Problem Sizes and Solutions

unzip posts.zipfor the zip file or

zcat posts.tar.Z | tar -xvf -Instructions and solution values are in README files in each package. Go Back to Contents

- Tolerances and solutions
- All stages should be solved to a relative tolerance of 10^(-6), i.e. if UB and LB are upper and lower bounds on the recourse objective in any given stage, (UB-LB)/(|LB|+0.1) <= 10^(-6). Interior point codes should use the same tolerance, but are not required to provide a basic solution.
- Primal and dual solutions should be given for all nodes in the solution tree. (If not, at least primal solutions should be given, and the absence of dual solutions should be reported.)

- Reporting
- Complete descriptions of the hardware and software used should be given, including machines make, model, memory, and speed (if applicable). If parallel codes are used, some indication of processor-processor bandwidth and message latency should be reported. (Ideally, the guidelines presented in Section 7 of Barr and Hickman should be followed.) If external routines are used (e.g. LP solvers), they should be cited.
- Both algorithmic solution times and total times should be reported. Algorithmic solution times exclude input, problem set-up, and output. System times should be used if uniprocessor codes are used. If a parallel code is being used, the best and average wall clock times over a group of tests should be used. Using machines loaded with other processes should be avoided, if possible.

Stages Scenarios Cor Time STOCH --------------------------------------------------------------- 3 25 sgpf3y3.cor sgpf3y3.tim sgpf3y3.sce 4 125 sgpf3y4.cor sgpf3y4.tim sgpf3y4.sce 5 625 sgpf3y5.cor sgpf3y5.tim sgpf3y5.sce 6 3125 sgpf3y6.cor sgpf3y6.tim sgpf3y6.sce

Stages Scens/stage Cor Time STOCH ---------------------------------------------------- 2 6 fxm.cor fxm2.tim fxm2_6.sto 16 fxm2_16.sto 3 6 fxm.cor fxm3.tim fxm3_6.sto 16 fxm3_16.sto 4 6 fxm.cor fxm4.tim fxm4_6.sto 16 fxm4_16.stoNote: The period partitions in the core file are different from those originally used in Gassman 1988. The original fxm file from the netlib test set was partitioned so as to have roughly equal size blocks.

Stages Scens/stage Cor Time STOCH --------------------------------------------------------------- 2 6 pltexpA2.cor pltexpA2.tim pltexpA2_6.sto 16 pltexpA2_16.sto 3 6 pltexpA3.cor pltexpA3.tim pltexpA3_6.sto 16 pltexpA3_16.sto 4 6 pltexpA4.cor pltexpA4.tim pltexpA4_6.sto 16 pltexpA4_16.sto 5 6 pltexpA5.cor pltexpA5.tim pltexpA5_6.sto 16 pltexpA5_16.sto 6 6 pltexpA6.cor pltexpA6.tim pltexpA6_6.sto

Stages Scens in Stg 2 Cor Time STOCH --------------------------------------------------------------- 3 6 pltexpA3.cor pltexpA3.tim pltexpB3_6.sto 4 6 pltexpA4.cor pltexpA4.tim pltexpB4_6.sto 5 6 pltexpA5.cor pltexpA5.tim pltexpB5_6.sto

Stages Scenarios Cor Time STOCH ------------------------------------------------------------- 2 8 stormG2.cor stormG2.tim stormG2_8.sto 27 stormG2_27.sto 125 stormG2_125.sto 1000 stormG2_1000.sto

- J. R. Birge, M.A.H. Dempster, H. I. Gassman, E. A.
Gunn, A. J. King, and S. W. Wallace, 1987. ``A Standard forinput format for
multiperiod stochastic linear programs,''
*COAL newsletter*, 17, pp. 1-20. - K. Frauendorfer,
Lecture Notes in Economics and Mathematical Systems No 392,
**Stochastic Two-Stage Programming**(Springer-Verlag, Berlin, 1993). - H. I. Gassman, 1990. ``MSLiP: A computer code for the
multistage stochastic linear programming problem.''
*Mathematical Programming*, 47, pp. 407-423. - J. K. Ho, 1975. ``Optimal Design of Multistage Structures:
A Nested Decomposition Approach,''
*Computers and Structures*, Vol. 5, pp 249-255. - M. J. Sims, 1992. ``Use of a stochastic capacity planning model to find the optimal level of flexibility for a manufacturing system,'' Senior Design Project, Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, MI 48109.
- J. M. Mulvey, and A. Ruszczynski, 1992. ``A New Scenario Decomposition Method for Large Scale Stochastic Optimization,'' Technical Report SOR-91-19, Dept. of Civil Engineering and Operations Research, Princeton Univ. Princeton, N.J. 08544

Det. Equiv. Problem Stages Scen/Stage Rows Columns Total Scen. Opt. Obj. Value ------------------------------------------------------------------------- PLTEXP pltexpA2_6 2 6 686 1856 6 -9.479354 pltexpA2_16 2 16 1726 4636 16 -9.663308 pltexpA3_6 3 6 4430 11864 36 -13.969368 pltexpA3_16 3 16 28350 75804 256 -14.267458 pltexpA4_6 4 6 26894 71912 216 -19.599417 pltexpA4_16 4 16 454334 1214492 4096 -18.849337 pltexpA5_6 5 6 161678 432200 1296 -23.214073 pltexpA6_6 5 6 970382 2593928 7776 -28.134408 pltexpB3_6 3 6+1+1 6 -13.643226 pltexpB4_6 3 6+1+1+1 6 -17.928191 pltexpB5_6 3 6+1+1+1+1 6 -23.846166 ------------------------------------------------------------------------- SCFXM fxm2.6 2 6 780 1047 6 18416.686 fxm2.16 2 16 1680 2227 16 18416.655 fxm3.6 3 6 4020 5295 36 18615.932 fxm3.16 3 16 24720 32435 256 18438.891 fxm4.6 4 6 23460 30783 216 18616.224 fxm4.16 4 16 393360 515763 4096 18438.891 ------------------------------------------------------------------------- STORM stormG2.8 2 8 2985 11456 8 15535231.897 stormG2.27 2 27 9635 37809 27 15508982.306 stormG2.125 2 125 43935 173735 125 15512090.180 stormG2.1000 2 1000 350185 1387360 1000 15802589.698 ------------------------------------------------------------------------- SGPF sgpf3y3 3 5 1220 1595 25 -2967.917 sgpf3y4 4 5 6097 7974 125 -3994.198 sgpf3y5 5 5 30487 39868 625 -5172.165 sgpf3y6 6 5 152434 199341 3125 -6463.323 sgpf5y3 3 5 1282 1342 25 -3027.706 sgpf5y4 4 5 5657 5726 125 -4031.391 sgpf5y5 5 5 24407 24476 625 -5201.282 sgpf5y6 6 5 246077 308733 3125 -6479.614Go Back to Contents

This page was last modified on . Send questions to jrbirge@northwestern.edu.