Well, the performance of algorithms like GP depend much on empirical studies and observation of results. Normally people have 'gut-feelings' about their approaches. Then they implement them and test as to how they perform on a spectrum on problems. So it may be worth trying this. It may as well be possible to try it out to see how it performs. A very simple alternative could be to restrict your function and terminal set to the ingredients of 'your' function. Let GP initialize a population, and you may see how things proceed from there on. This way you may ensure that the population has enhanced diversity by remaining in the 'concerned search space' and also overcome the problem of stagnation as well.

It may be worthwhile to read some relevant literature.


----- Original Message ----
From: Shane <[log in to unmask]>
To: [log in to unmask]
Sent: Friday, April 25, 2008 9:43:45 PM
Subject: Re: Initialize trees

Ok.  I was thinking I would initialize all of the individuals with this original tree and then mutate all of them slightly, so they would all vary somewhat from the original.  Then follow the normal process after 'inittree'.  Do you think this would work?

Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now.