After reading the responses and pondering the problem, I realize that I was
almost certainly posing the wrong question.
I can produce a solution in multiple passes by reformulating the problem
between passes.
To wit:
If sum{j in J} B[i,j] = 0 drop all of the ith terms and seek a solution to the
new set of equations.
That's at worst clunky, but clearly doable. So it seems to me the concept is
sound even though my expression of it was not. I am not able to say if it's
doable in a single pass. This is very much of a learning experience for me.
I'd be very interested in seeing the binary variable example from p. 43 of the
CMPL manual in GMPL form. Or any other examples that impose the sort of logic
outlined above. It may not help me now, but might prove useful later. The
disjoint feature sets and notation of AMPL, CMPL, GMPL, etc. make things
challenging for a novice reading everything he can get a hold of.
Thanks to all,
Reg
_______________________________________________
Help-glpk mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/help-glpk
