Here is an example of how to get the problem "file name" from the
directory you are working in type:
cp/home/or641/or641/lp/"file name"
Note : To obtain a directory of your files type: ls
To obtain a detailed directory ( including date of creation, size,
etc. ) of your files type ls -l
Some information about this problems: fit1p and fit1d are in a sense,
primal and dual versions of the same problem ( the dual was transformed
to a minimization ). They originate from a model for finding a set of weights
that can be used to score credit applications; the problem israel is a
sparse problem with a few dense columns. As you will see, this can be challengind
for some of the methods.
Some of the problems in this set may be degenerate.
To run cplex you must telnet site.gmu.edu then type
/opt/cplex/cplex
you must first read the problem,say
read fit1p.mps
then to run the simplex,type
opt
(short for optimize), or if you want the dual simplex type
tran
(short for tranopt), or if you the interior point algorithm type
baropt
If you run a problem twice, the second time it will use last available
solution. If you want to rerun it, you should reread it.
The display command will give you various display (output) options.
Type display to see them. Display stat gives you information
about the problem statistics.
The help command will give you the major cplex commands.If you
want to change some parameters, use the set command---it will prompt
you for the changes. For the simplex parameters, type set simplex
The goal of he project is to compare different LP algorithms with different
settings, for solution of these five problems.
Run each of the problems
(a) with the simplex method
(b) with the dual simplex method
(c) with barrier method
In each case run the problem with its default settings, but
( i ) preprocessing ( the default; to get it type default )
( ii )with no preprocessing. ( in set preprocessing ,set aggregator,
coeffreduce and presolve all to " no " ).
For the primal simplex and the dual simplex ( with the default preprocessing
), test also the effect of different pricing options. To change the pricing
option for the primal algorithm type
set simplex pgradient
and choose options 0 , 2 , and 4. To change the pricing option for
the dual algorithm, type
set simplex dgradient
and choose options 0 , 1 , and 2. ( we will discuss the options in
class).
Summarize your results in one or more table for each problem. State
the problem size, and the noumber of non zeros, before and after preprocessing.Also
summarize the results of each run, in terms of total solution time, tyhe
number of iterationd, and average time per iteration. Note that the barrir
method crosses over at the end to the simplex method. Provide information
about each of the stages individually, as well as total time.
Give a short summary of the results of the runs. Include any conclusions that you can draw from these runs.