Question about fitness

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Question about fitness

Applegate, Doug
Dear Klaus:
 
I have a question regarding a comment on the abstract class FitnessFunction. It says:
 
Note: Two Chromosomes with equivalent sets of genes should always be assigned the same fitness value by any implementation of this interface.
 
On my particular implementation, I am analyzing Neural Networks. I'm using the JOONE package, I'm not sure if you are familiar with that or not. Actually, I noticed that there was a dormant project called JooneGap that was involved in evolving networks, but it looks pretty old, I'm not sure if anyone is active on it. Anyhow, my question is, why is it important that 2 chromosomes return the same fitness value? Due to the random initial conditions of a neural network, i.e. the starting weights, 2 networks with exactly the same parameters will result in different outcomes, and thus different fitnesses. So, I am just curious about the impact on the GA given this situation. Thank you in advance for your help.
 
Regards,
 
Doug

Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642

_______________________________________________
jgap-devl mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jgap-devl
Reply | Threaded
Open this post in threaded view
|

Re: Question about fitness

Klaus Meffert-2
Dear Doug,
 
you are right about JOONEGAP, it has not been maintained in the past, although it should.
Regarding your original question: Two chromosomes equivalent concerning their genes should get the same fitness value assigned because they are "equal". Two twins are "equivalent" in the same sense. It has nothing to do with neural networks, it's a statement about an evolutionary principle: assign two "equal" individuals the same weights, that's all.
 
That there is a NN behind the GA is a different thing that is not correlated to the statement with the 2 chromosomes. But I don't remember specific details about JOONEGAP although I worked on that project a while ago.
 
Best
 
Klaus


From: [hidden email] [mailto:[hidden email]] On Behalf Of Applegate, Doug
Sent: Tuesday, June 27, 2006 3:20 PM
To: [hidden email]
Subject: [jgap-devl] Question about fitness

Dear Klaus:
 
I have a question regarding a comment on the abstract class FitnessFunction. It says:
 
Note: Two Chromosomes with equivalent sets of genes should always be assigned the same fitness value by any implementation of this interface.
 
On my particular implementation, I am analyzing Neural Networks. I'm using the JOONE package, I'm not sure if you are familiar with that or not. Actually, I noticed that there was a dormant project called JooneGap that was involved in evolving networks, but it looks pretty old, I'm not sure if anyone is active on it. Anyhow, my question is, why is it important that 2 chromosomes return the same fitness value? Due to the random initial conditions of a neural network, i.e. the starting weights, 2 networks with exactly the same parameters will result in different outcomes, and thus different fitnesses. So, I am just curious about the impact on the GA given this situation. Thank you in advance for your help.
 
Regards,
 
Doug

Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642

_______________________________________________
jgap-devl mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jgap-devl