[galib] Quick question on mutation probability
Dr. Ian D. Wilson
ian.d.wilson at btinternet.com
Fri Jun 11 04:00:08 EDT 2004
Camilo,
The change occurs in probability, which means that it isn't guaranteed to be
1% all the time, just that on average it will be 1%.
Some researchers have published "rules of thumb" for choosing the best
mutation rate based on the length of the chromosome and the population size.
DeJong suggested that the mutation rate be inversely proportional to the
population size. Hessner and Manner suggest that the optimal mutation rate
is approximately:
(M * L1/2)-1
where M is the population size and L is the length of the chromosome.
Of course, you cannot expect heuristics to be perfect but this can be
considered to be indicative.
I hope that helps shed some light.
Best wishes,
Ian
De Jong, K.A. Ph.D. Dissertation, University of Michigan, 1975. (GA
configuration)
Schaffer, J.D.; Caruna, R.A.; Eshelman, L.J.
Hessner, J.; Männer, R. in Proceedings of the First Workshop on Parallel
Problem Solving from Nature (Lecture Notes in Computer Science, Vol. 496);
P. Schwefel and R. Männer (Eds.); Springer-Verlag: Berlin, 1991, pp 23-31.
(choosing optimal mutation rates)
-----Original Message-----
From: galib-bounces at mit.edu [mailto:galib-bounces at mit.edu]On Behalf Of
Camilo Gomez
Sent: 10 June 2004 19:44
To: galib at mit.edu
Subject: [galib] Quick question on mutation probability
All,
If pMut = 0.01 and my binary string has a length of 200 bits. That means
that at least two bits are always mutating on every chromosome?
There is no chromosome that exhibits no change?
Isn't that bad?
If I am correct, the pMut should be proportional to the string size.
Right?
Any comments?
Thanks,
Camilo Gómez
"About 60 million transistors were built this year just for you, with
another 60 million for each of your friends, plus 60 million for every
other man, woman, and child on Earth. By 2010, the number should be
around one billion transistors per person per year. "
http://www.embedded.com/story/OEG20021217S0039
-----Original Message-----
From: galib-bounces at mit.edu [mailto:galib-bounces at mit.edu] On Behalf Of
John Conner
Sent: Monday, May 17, 2004 3:50 PM
To: galib at mit.edu; Laurent Steffan
Subject: Re: [galib] Multiobjective (pareto) optimization?
I don't know if this will suit my needs, but I'm very willing to look
at it tomorrow. Thanks greatly for suggesting it!
- John
On May 17, 2004, at 5:28 PM, Laurent Steffan wrote:
> Hello,
>
> If you're in a hurry, you could also have a look at "Open BEAGLE", a
> C++ GA (and others) framework, with a nice documentation, that offers
> multi-objective optimization. In fact one of the examples is a
> two-objective optimization of a knapsack problem, which generates a
> nice Pareto front. Does that look like anything you need ?
>
> http://www.gel.ulaval.ca/~beagle/
>
> BTW, I'm just a user of Open Beagle and not in any other way connected
> with it :-)
>
> --
> Laurent Steffan, Ph. D.
> Scientific Consultant
> +33 (0) 6 85 92 08 51
>
> On 17.05.2004 14:39, John Conner wrote:
>> Please! I'm trying to use GAlib to produce some publishable work, and
>> I really need the multi-objective functionality to make a good show,
>> so if you can send me anything, the sooner the better! Thank you very
>> much.
>> - John
>> On May 17, 2004, at 4:53 AM, GAAL Balazs wrote:
>>> Hi!
>>>
>>> We are working on a multi-objective, multi-level extension to GALib.
>>> Hopefuly it will be useable in June or July.
>>> It is built on top of GALib. Our fitness implementation classifies
>>> genomes according to numerical constraints and rules. This is done
>>> in a multi-objective way.
>>>
>>> If you are interested, I could send you some source code or the
>>> description of our fitness function and genome coding.
>>>
>>> GAAL, Balazs (M.Sc.)
>>> PhD Student
>>> Dept. of Information Systems
>>> University of Veszprem
>>> HUNGARY
>>>
>>> John Conner írta:
>>>
>>>> I've searched the back archives for this list and have found some
>>>> messages asking about multiobjective optimization with GALib. I've
>>>> also found no solutions, so I'm hoping that someone is paying
>>>> attention to this wacky thing and can let me know of a good option
>>>> for handling multiobjective optimization with a library. I can
>>>> switch from GAlib to another c++ library if necessary, though if it
>>>> is easy enough to extend GAlib to do the multiobjective stuff, I
>>>> would like to keep using it.
>>>>
>>>> Has anyone done this? Has anyone tried and failed? Please get back
>>>> to me!
_______________________________________________
galib mailing list
galib at mit.edu
http://mailman.mit.edu/mailman/listinfo/galib
_______________________________________________
galib mailing list
galib at mit.edu
http://mailman.mit.edu/mailman/listinfo/galib
More information about the galib
mailing list