[Coin-symphony] Sensitivity Analysis

[David Goodlad]

Hi Menal,

I fixed the ub function as well, and added a check for the objective
sense (since I'm doing maximization).  It didn't make much sense to me
to have to pass in negative objective function coefficients here,
while I could use positive values elsewhere :)  I also wasn't too
happy with the function assuming that any coefficients that I didn't
explicitly pass in as non-zero would be equal to their values in the
original function.  I wrote two versions, one that maintained the same
logic (so that you could pass in, say, only 2 variables and have the
rest of the function remain the same) and another that ignores the
original function and only calculates values based on the coefficients
passed in.  I'd be happy to provide this updated function, but am
unsure in what format you would want it.  'cvs diff' is kind of ugly,
and I forget the typical options for 'diff' :)

I'll probably look into implementing the lower-bound function at some
point, but first I have to run some more time-consuming experiments to
see if I'll even need it...

Dave

On 2/7/06, Menal Guzelsoy <megb at lehigh.edu> wrote:
>
> Dave,
>
> The usage of those variables does not depend on the status of the node. I have
> just modified that function with a fix to those lines. It seems to be working
> now.
>
> We are planning to get to the function 'get_lb_for_new_obj' soon, however
> you are always welcome to contribute to the project if you feel to
> implement it
> before we do.
>
> Menal.
>
> Quoting David Goodlad <dgoodlad at gmail.com>:
>
> > Hi again
> >
> > I've been playing with the sym_get_ub_for_new_obj function, and have
> > found that it's not working properly.  It segfaults when building the
> > new objective value...
> >
> > Anyway, from some analysis, it seems that the function
> > get_ub_for_new_obj is very broken.  I'm working on fixing it (I'll
> > supply a patch once I get it going), but have one question.  The
> > sol/sol_ind/sol_cnt variables, making up the solution in the bc_node
> > struct, seem to store the values/indices of the non-zero variables in
> > the solution.  However, there are spots in the current function that
> > refer to it in a different way (see lines 3044-3045 compared to lines
> > 3055-3056 in master_func.c).  Does the usage of these variables change
> > depending on feasibility status of the node?  I wouldn't expect it to,
> > but I only started looking at Symphony last weekend, so can't be sure!
> >
> > Also, is there a planned (or just uncommitted) version of
> > get_lb_for_new_obj?  Otherwise, I am going to look into implementing
> > that as well.
> >
> > Thanks
> > Dave
> >
> > On 2/7/06, David Goodlad <dgoodlad at gmail.com> wrote:
> >> Hi Menal,
> >>
> >> Thanks for the quick answer.  It compiles fine now.
> >>
> >> Dave
> >>
> >> On 2/7/06, Menal Guzelsoy <megb at lehigh.edu> wrote:
> >> >
> >> > David,
> >> >
> >> > That function is not complete yet and needed to be commented out.
> >> Just do an
> >> > update now.
> >> >
> >> > Menal.
> >> >
> >> > Quoting David Goodlad <dgoodlad at gmail.com>:
> >> >
> >> > > Hi all
> >> > >
> >> > > It seems that when building Symphony with SENSITIVITY_ANALYSIS enabled
> >> > > in the config, I end up with a library missing a function definition.
> >> > > get_lb_for_new_obj is declared in the master.h file but looking into
> >> > > master_func.c where similar functions are defined shows that it's just
> >> > > 'missing'.  Is this functionality just incomplete for now, or am I
> >> > > missing something here?
> >> > >
> >> > > This is with the latest from CVS, by the way.
> >> > >
> >> > > Thanks!
> >> > > Dave
> >> > >
> >> > > --
> >> > > Dave Goodlad
> >> > > dgoodlad at gmail.com or dave at goodlad.ca
> >> > > http://david.goodlad.ca/
> >> > >
> >> > > _______________________________________________
> >> > > Coin-symphony mailing list
> >> > > Coin-symphony at list.coin-or.org
> >> > > http://list.coin-or.org/mailman/listinfo/coin-symphony
> >> > >
> >> >
> >> >
> >> >
> >> > ----------------------------------------------------------------
> >> > This message was sent using IMP, the Internet Messaging Program.
> >> >
> >> > _______________________________________________
> >> > Coin-symphony mailing list
> >> > Coin-symphony at list.coin-or.org
> >> > http://list.coin-or.org/mailman/listinfo/coin-symphony
> >> >
> >>
> >>
> >> --
> >> Dave Goodlad
> >> dgoodlad at gmail.com or dave at goodlad.ca
> >> http://david.goodlad.ca/
> >>
> >
> >
> > --
> > Dave Goodlad
> > dgoodlad at gmail.com or dave at goodlad.ca
> > http://david.goodlad.ca/
> >
>
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
>


--
Dave Goodlad
dgoodlad at gmail.com or dave at goodlad.ca
http://david.goodlad.ca/