nataraj chakraborty wrote:
> Mitchell,
> Well as I explained if you penalise those chrome which performs not
> similar for all  the 32 patterns and encourage those whose devation
> from the mean is less that is the performance is similar and falls
> within a range of values over all the 32 patterns than the results may
> be better....
> You may try this.... in next few evolution....
>
> Algo may be something like this....
>
> Store all the fittness(Performance) values for all 32 patterns in a
> sorted array....
> The middle of the array is the mean....
> Now find the largest difference with the mean of any of the above
> elements in the array...
> If the largest value is < than a speciefied constant.... than the
> fittness is inversely proproiotnal to the largest diffrence....
>   
I don't really understand what you are suggesting, but you should know 
that there are about 500 patterns.  The 32 patterns was when we were 
using exactly 5 steps.  There are 32 patterns for 5 steps.   Now, we are 
using all path lengths from 3 to 8 steps.  8 steps has 256 patterns, 7 
has 128, etc.
> ***
> This will encourage the ANN to perform in a similar manner for all the
> range of i/p's
> also Evoltuion may be tougher in that case.....
> ***
> I think you do penalise the Chrome based on no. of updates.....
> The point  is to encourage those Chrome requiring only one update.....
> ***
> The above two actually ellucidates a practical scenerio.....
> *******************************************************************************
>
> May we should take up all these (the above Ideas).... only after we'r
> finished with our present crusade!!!
>   


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

Reply via email to