Re: [eclipse-clp-users] Pareto Optimality + Branch and Bound?

From: Marco Zimmerling <marco_at_...260...>
Date: Thu, 29 Apr 2010 10:38:20 +0200
Hi Philipp,

I'm not aware of any ECLiPSe framework for Pareto front 
generation/approximation.

As Andrzej suggested, one of the most common methods is the weighted sum 
approach. Assuming two objective functions f1(X) and f2(X), with X being 
the vector of decision variables, the single criteria problem becomes

   f(X) = w1*f1(X) + w2*f2(X),

where w1 and w2 are weights that *should* reflect the importance of each 
objective. Typically, one assumes w1 + w2 = 1. By solving this problem 
for different combinations of w1 and w2, you can generate different 
Pareto optimal points. However, it is impossible to obtain points on 
non-convex portions of the Pareto front, and it is often difficult to 
generate a smooth Pareto front (i.e., without holes). So it is 
worthwhile to look into other approach as well. The article "Survey of 
multi-objective optimization methods for engineering" by Marler and 
Arora (Struct Multidics Optim 26, pp. 369-395, 2004) provides a good 
overview of multiobjective optimization methods, including a discussion 
of pros and cons.

Another branch of research uses evolutionary methods to find good 
approximations of the Pareto front. The optimization problem is treated 
as a black box; the focus lies on efficient stochastic search methods. 
If you want to get a feeling: 
http://www.tik.ee.ethz.ch/sop/publicationListFiles/zlb2004a.pdf

Marco

Andrzej Lewandowski wrote:
> Convert to standard optimization problem (i.e. single criteria) using
> weighted sum of individual criteria
> 
> A.L.
> 
> -----Original Message-----
> From: Philipp Marcus [mailto:marcus@...247...] 
> Sent: Wednesday, April 28, 2010 4:13 PM
> To: eclipse-clp-users_at_lists.sourceforge.net
> Subject: [eclipse-clp-users] Pareto Optimality + Branch and Bound?
> 
> Hi,
> 
> I have a COP with two cost variables and I'm searching for a pareto 
> optimal solution. Though it would be nice to get the complete pareto 
> front, I'd already be happy with *one* pareto optimal value. Is there a 
> framework I could use? If not so, do you have any suggestions how this 
> could be implemented?
> 
> Best regards,
> Philipp
> 
> ----------------------------------------------------------------------------
> --
> _______________________________________________
> ECLiPSe-CLP-Users mailing list
> ECLiPSe-CLP-Users_at_lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/eclipse-clp-users
> 
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> ECLiPSe-CLP-Users mailing list
> ECLiPSe-CLP-Users_at_lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/eclipse-clp-users
> 
Received on Thu Apr 29 2010 - 08:56:12 CEST

This archive was generated by hypermail 2.2.0 : Mon Jul 09 2018 - 02:05:29 CEST