[Octave-bug-tracker] [bug #66010] Function glpk produces incorrect output

[anonymous]

Follow-up Comment #12, bug #66010 (group octave):

Reacting to comment #11:
"Or for contraints matrix A with such dodgy values (< eps) presolving converts
obviously the original (unsolvable) constraint matrix to a solvable albeit
incorrect "equivalent" one"

I strongly doubt this is the case here. I mean that I believe the original
constraint matrix in my data files which includes the small values < eps() is
in fact solvable. This is why it did not occur to be that these small values
could be a problem.

In general, these LP problems can be very sensitive to the data. For example,
a constraint with an equality a*x==b is formally equivalent to two constraints
with inequalities a*x<=b and a*x>=b but here obviously small changes in the
values can make the problem unsolvable or have huge effects on the solution.

This being said, I don't think my example is of this type, which is why it is
surprising that cleaning the data has such a huge effect.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?66010>

_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/

signature.asc
Description: PGP signature