From: Amine Marref <marref_at_cs.york.ac.uk>

Date: Wed, 02 Jan 2008 15:45:27 +0000

Date: Wed, 02 Jan 2008 15:45:27 +0000

Hi Folks, Can anybody spare some time to answer this very question. I use Eclipse as a constraint solver in my research. I am not sure though that this is the right investment. The problems I model *all* have the following properties. * All variables are integers. * At least two thirds (2/3) of the variables have domain [0..1]. * Each variable in the other remaining third (1/3) of variables can take a finite number of different values only, usually less than 20. * The number of variables is in tens of thousands (50000 variables is the average). * All constraints are linear: addition, subtraction, and implication are used. * The objective function is non-linear, expressed as: (Obj=B1*C1 + B2*C2 + ... + Bn*Cn) where Bi, Ci are integer variables. Currently I use bb_min(search(Vars,0,input_order,indomain_split,complete,[]) from "branch_and_bound" to do the search (Vars is the list of all variables in the program). The search takes ages. Any suggestions for improvement? Or is there a more appropriate constraint solver for this? Thanks a lot, Amine. -- Amine Marref PhD Research Student Real-Time Systems Group Department of Computer Science The University of York Heslington YO10 5DD York United Kingdom +44 1904 432810Received on Wed Jan 02 2008 - 15:48:19 CET

*
This archive was generated by hypermail 2.2.0
: Thu Feb 02 2012 - 02:31:58 CET
*