home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ukma!gatech!pitt.edu!genetic
- From: genetic+@pitt.edu (David M. Tate)
- Newsgroups: comp.ai.genetic
- Subject: Re: Penalty functions
- Keywords: Penalty Optimization Minimization
- Message-ID: <2716@blue.cis.pitt.edu>
- Date: 28 Jan 93 16:23:13 GMT
- References: <C1F8H9.8nw@world.std.com> <1853@airgun.wg.waii.com>
- Sender: news+@pitt.edu
- Organization: Department of Industrial Engineering
- Lines: 24
-
- hart@de01.denver.waii.com said:
- >
- >Multiplier methods, for constrained optimization problems, generally
- >exceed the performance of penalty function methods. The reference:
- >R.A. Tapia, "Diagonalized Multiplier Methods and Quasi-Newton Methods for
- >Constrainded Optimization", Journal of Optimization Theory and Applications,
- >Vol. 22., No.2, June 1977, p. 135-194., is where I got started with these
- >methods. Two papers on Penalty Function methods are in the reference section
- >of this paper. They should point you to some examples of penalty functions.
-
-
- Multiplier methods are fine for *continuous* optimization, but not readily
- applicable for many combinatorial problems, unless a convenient continuous
- relaxation exists. Multiplier methods also require that you have an explicit
- representation of the constraint set in a particular form. For example, in
- a Travelling Salesman problem the logically simply constraint "Each tour must
- pass through each city exactly once" translates into an exponential number of
- linear inequalities in the linear/integer programming formulation.
-
- --
- David Tate |"Something there is that doesn't love a wall,
- dtate@unix.cis.pitt.edu | That sends the screaming liner over it
- Prof. of Story Problems | And spills the bleacher-dwellers from their seats."
- Member ORSA, TIMS, SABR | "Fenway Wall", Robert Frost
-