[Coin-bcpdiscuss] Both cut and variables generation

[Laszlo Ladanyi]

Hi Tung,

Unfortunately you cannot generate both cuts and variables in the same
iteration, and the reason is exactly what you describe. In each iteration you
got to decide which one you generate. 

It'd be possible to extend bcp to handle generating both, but it would have
other undesirable consequences, like if in an iteration you generate only cuts
then you know the objective value can only increase, and if you generate only
vars then the objective can only decrease. That information might come handy.
Also, a cut that was violated now may not be violated at all if you include
the vars generated in the same iteration and reoptimize. Or vice versa, a
generated var may not have negative reduced cost if cuts generated in the same
iter would have been included in the optimization. I think it's just cleaner
to generate only one kind in any given iteration...

Sorry for the bad news and I hope my reasons for not implementing what you'd
like are convincing enough. Still if you believe the ability would help for
your problem and you want to add the option then go ahead and I'll merge your
changes to the base code.

Cheers,
--Laci

On Wed, 2 May 2007 tle at purdue.edu wrote:

> Hi,
> 
> My program generates new cuts and variables at each pricing iteration.
> Therefore, i need to use both " vars_to_cols" and "cuts_to_rows" at same time.
> Problem is that:
> 
> - "vars_to_cols"  works for active vars and old cuts ( not including active 
> cuts)
> - "cuts_to_rows" works for active cuts and old variables ( not including 
> active vars)
> 
> Therefore, i dont' know how to define the coefficients intersecting by active 
> cuts and active vars.
> 
> Pls help me
> Thanks
> Tung Le
> PhD student
> Purdue University
> 
> 
> _______________________________________________
> Coin-bcpdiscuss mailing list
> Coin-bcpdiscuss at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-bcpdiscuss
>