[Coin-discuss] setColSolution

[Mikhail Nediak]

A question: would it be correct to say that setColSolution() was planned
primarily as a way to set warm start information for interior-point type
methods?

If the answer is yes, it is not surprising that simplex type methods
often have some trouble interpreting it. If a passed solution has
enough variables at their bounds, one could force it on the solver
by a big-M method. It would not work for all cases though and I am not
sure it would be worthwhile computationally to force this kind of warm
start on a simplex solver. Another possibility then is to define a
CoinWarmStartColSolution descendant of CoinWarmStart and pass a
column solution to the solver in this way.

Best,
Mikhail

On Tue, 29 Apr 2003, Lou Hafer wrote:

>
> 	Brady writes, in regard to set[Col,Row]Solution for glpk,
>
>   One option is to "set" the values that are cached, but not tell GLPK
>   about them at all.
>
> 	For what it's worth, this is the approach I'm taking with dylp, with
> a warning in the documentation that dylp will take no notice of these
> values.  This seems to match the approach in clp (as best I can figure; fun
> to read but John has too much time on his hands, or maybe too many years of
> practice :-).
>
> 	It really seems to me that the utility of these two methods is to
> allow the user to establish useful values for the solutions and objective
> prior to calling the solver. If you really want to force a solution on dylp,
> the proper approach is to give it a basis and a status vector (likely true
> for most simplex solvers).  My impression is that this is handled through the
> OsiSimplex interface, but I haven't quite gotten there yet.
>
> 	This is one of the ambiguities that prompted me to suggest that we
> need to work up some underlying principles for our interface layers.
>
> 						Lou
>
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at www-124.ibm.com
> http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-discuss
>