Solving Two-Stage Linear Stochastic Optimization Problems with Recourse on Computational Grids
This site contains codes and reports relevant to our ATR solver
for two-stage stochastic linear programming problems with recourse.
The solver uses the
MW toolkit
to run as a master-worker algorithm on a
variety of distributed computational platforms. Most of our
computational experience has been with the
Condor system's
implementation
of MW.
ATR was developed as part of the
metaneos project.
Technical Reports
A comprehensive description of the algorithm and its implementation:
-
J. Linderoth and S. J. Wright,
"Decomposition algorithms for stochastic programming on a computational grid,"
Preprint ANL/MCS-P875-0401, April, 2001:
pdf.
A summary paper:
-
J. Linderoth and S. J. Wright,
"Computational Grids for Stochastic Programming,"
Optimization Technical Report 01-01, Computer Sciences Department,
University of Wisconsin-Madison, October, 2001; revised February, 2002:
pdf.
A report on empirical experiments performed with the code (see also
this web site):
-
J. T. Linderoth, A. Shapiro, and S. J. Wright, The Empirical
Behavior of Sampling Methods for Stochastic Programming,
Optimization Technical Report 02-01, Computer Science Department,
University of Wisconsin-Madison, January, 2002:
postscript,
pdf.
Codes
- ATR Source
- SUTIL: Tool for generating sample-average approximations to a
two-stage stochastic linear program.
- LPSOLVER: Unified callable-library interface to various linear
programming solvers.
The MW libraries
must also be installed, together with a computational grid
infrastructure on which MW can execute, such as
Condor.
Links
Here are some links to sites with computational stochastic programming
data, in various states of repair.
Support
The Department of Energy, Office of Advanced Scientific Computing
Research
The National Science Foundation Grant CDA-9726385.