[gecode-users] Dynamically constructing a model using solution of an IntVar

Ajit Dingankar Ajit.Dingankar at gmail.com
Thu Aug 15 05:33:22 CEST 2013


> That's more difficult.  Let's say your variable that contains the required 
array length is called x.  You
> either need to decompose your model into two parts (the first part 
determining x, the second part then
> using that solution to construct a new model with the right length), or 
you need to allocate an array with
> the maximum possible length and use constraints to say that everything 
beyond index x is ignored.
> 
> Cheers,
> Guido

Thanks, Guido! I was thinking about an approach similar to your first 
suggestion which may be better suited for my situation (in addition to 
being more memory efficient ;-) since it's a natural fit for my solution 
requirement: I need to solve for every legal length but the data should 
be randomly picked (one solution for every length) since the data space 
is very large and most solutions don't add any insight into the problem 
over that of a single representative solution. 

I couldn't find an example of model decomposition in MPG, so I'd really 
appreciate any pointers or hints of handling the underlying space; mainly 
if I can/must use a different space for the second model.  I think I've 
to use a different space and the variables in the first model can't be 
constrained by those in the second model in this approach. If there are 
any such constraints, I must use the second approach of maximum possible 
length. 

Thanks, 
Ajit
==== 





More information about the users mailing list