I've just started trying Ben Sapp's lp.cc code (from octave-forge) on
Mac OS X, and I'm getting some strange results. Before I dig too
deeply, I'd like to see if these results occur on other platforms as
well or if this is unique to the (currently experimental) .oct file work
on Mac OS X.
Here's the m-file I use to test lp: all solutions should lie within the
range [-10,10], with the number of equality constraints increasing from
0 to two as the iteration goes. The results on my machine are at the
end. Notice that the upper bound is violated and that the equality
constraints are not enforced.
If someone would run this script and let me know if my results are
duplicated, I'd appreciate it. If they are, then I'll debug the lp.cc
code itself. If not, then I've got something else going on that I'll
have to figure out.
Thanks.
A. S. Hodel
MyLpTest.m
============
A = eye(2,6);
c = [1 -2 3 -4 5 -6]';
b = [1;1];
lowerBound = -10*ones(size(c));
upperBound = 10*ones(size(c));
for nEqConstr = 0:2
printf("\n\nnEqConstr=%d\n",nEqConstr);
# compute results
x = lp(c',A,b,lowerBound,upperBound,nEqConstr);
# check against equality constraints, lower and upper bounds
mesg = str2mat("low"," ","high");
for ii=1:length(x)
istr = 2 - (x(ii) < lowerBound(ii)) + (x(ii) > upperBound(ii));
printf("%4d: %12.4g %12.4g %12.4g %s\n", ii, lowerBound(ii),
x(ii), ...
upperBound(ii), mesg(istr,:));
endfor
ConstraintCheck = A*x - b
end
========= Output of the program: =========
nEqConstr=0
1: -10 10 10
2: -10 21 10 high
3: -10 10 10
4: -10 30 10 high
5: -10 10 10
6: -10 30 10 high
ConstraintCheck =
9
20
nEqConstr=1
1: -10 10 10
2: -10 21 10 high
3: -10 10 10
4: -10 30 10 high
5: -10 10 10
6: -10 30 10 high
ConstraintCheck =
9
20
nEqConstr=2
1: -10 21 10 high
2: -10 21 10 high
3: -10 10 10
4: -10 30 10 high
5: -10 10 10
6: -10 30 10 high
ConstraintCheck =
20
20
_______________________________________________________________
Multimillion Dollar Computer Inventory
Live Webcast Auctions Thru Aug. 2002 - http://www.cowanalexander.com/calendar
_______________________________________________
Fink-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/fink-users
