galibuser

[B.ALI]

I am trying to GAlib for optemising neural network parameters or
weights. This for my Msc project.

The problem looks like this

suppose I have trained a neural network and following are the weights

w1,    w2 ,        w3,     w4  ,   w5 ,     w6
0.23   1.323     .221    .1123   1.2       1.1

I want to input above data as genome and produce next generation by
selecting  from say to points from above and mating them . Then isert
child into the population.

after one gen

0.23 ,  1.323   ,*1.113,.   1123   1 .2   1.1   

* assume that w4  and w3 produced this child

and then carry on this process.

aproximation function looks like this


a1*w1+a2*w2+a3*w3+a4*w4+a5*w5+a6*w6=f

obective finction look like this


score=1/fabs(f-p);

where p is the output we want, if the generation output "f" is very
colse to p then it returns higher score. The "wi" meby positive or
negetive. The initial w will be quit good but I want to improve on
that. 


I have tried to use several of your classes at the end I found that
GA1darray might be used. but I could not use it.


I want GAlib to use only above data for reproduction only.

I do not know how to use it.I have been working on it for over 2
months.

I have been working on it for over two months.
I have cosidered several classes but I think problem may lie in 
initializing the above genome. I want to treat the whole of the above
list as one genration. and produce another one from it. by selecting
inviduals from it.


Please reply if you can help. or need more information about the
problem situation. I have spent considerable amount of time on it.

Thank you



Bash
Msc Scientific Computing
c9503415 at hud.ac.uk