Mailing List Archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [tlug] genetic algorithm/optimization framework



Lars Kotthoff writes:

 > Right. The problem with applications usually is that it's not clear how the
 > different objectives should be weighted relative to each other :)

That almost goes without saying.  If you know the weights (or have a
more general functional aggregation of the objectives), you're back in
the single-objective world (mathematically -- in practical decision-
making, having the vector of sub-objective values can be extremely
valuable).

One thing I forgot to say (which the multiobjective packages do as a
matter of course) is to evaluate all of the objectives at each
solution.  That gives you a way to measure the cost (loss of value of
alternative objectives) of optimizing each objective while ignoring
the others.

This is especially important if some of the objectives are
"satisficing" (some level is "good enough" but you'll take
improvements if they're available).  It does occasionally happen that
no matter what you do the "good enough" level is achieved.  If not, it
may be a better approach to turn such objective into a constraint at
the "good enough" level.

 > Depending on the application, it may also make sense to have a
 > portfolio of different parameter settings instead of a single one,

You may also be able to kick the problem upstairs to your boss if you
have a bunch of solutions with different advantages.  Of course if
your boss is worth working for, she already knew that and that's why
you got delegated the problem in the first place! :-D

 > i.e. in one case you really care about bandwidth and want to use
 > the optimal configuration for that and in another case you care
 > more about something else. This would certainly make the
 > optimisation easier (and there are good methods for working with
 > such portfolios).
 > 
 > Of course it's also possible (albeit unlikely) that, when
 > optimising for each objective in turn, the same parameter settings
 > (or at least very close settings) come out at the top.

Isn't that the "everybody is welcome to my opinion" solution?  :-)



Home | Main Index | Thread Index

Home Page Mailing List Linux and Japan TLUG Members Links