The NEOS Server offers PENSDP (version 2.2) for the solution of semidefinite programming problems in
sparse SDPA format or in
SeDuMi format.
PENSDP is a computer program for solving problems of convex and
nonconvex nonlinear programming and (generally nonlinear)
semidefinite programming. Originally an implementation of the PBM
method of Ben-Tal and Zibulevsky for problems of structural
optimization, it grew up to a stand-alone program for solving general
problems. PENSDP is particularly aimed at large-scale problems
with sparse data structure. It is based on a generalized augmented
Lagrangian method.
More information on PENSDP is available in this
user's guide.
PENSDP was developed by
Michal Kocvara and
Michael Stingl.
It was implemented by
Hans Mittelmann and executes in 64 bits at
Using the NEOS Server for PENSDP
The user must submit a model in either sparse SDPA or SeDuMi Matlab
format to solve a semidefinite programming problem.
Examples of models in sparse SDPA format can be found in the
SDPLIB library.
The same problems in SeDuMi format are
here. Other files in this format are at
7th DIMACS Challenge library.
Note that when submitting via e-mail or XML-RPC empty
tokens need to be deleted.
If non-standard parameter settings are required, the user may also submit a
parameter file. You can check out the
default parameter file, edit it and resubmit it as part of the job.
In addition to PENSDP's own error output the 6 error measures are printed
according to the DIMACS 7th Challenge, see the
benchmarking paper.
This facilitates comparison with other SDP solvers.
|