Posted by Dan Carlson on December 03, 1997 at 18:45:57:
I have an overdetermined linear system I need to solve that
has a physical constraint imposed upon its solution. All of
the elements of the solution vector must be greater than zero.
Being a real-world problem, there is noise in the data that
cannot be eliminated. The simple approaches I have tried
for solving the linear system always result in at least a few
negative solution vector elements. The residual error is
sufficiently small, but the solutions are still unacceptable
because they violate the physical constraint.
Is this problem best handled as a linear system problem, or
as a linear programming problem? If it is best handled as a
linear system problem, then what is the best method of
Thanks for any assistance you can offer.