Print

Print


Hi,

Is there a known set of parameters for running the GP's sextic regression
problem so that it is finds a solution? I have tried running it multiple
times with the default parameters up to 2000 generations and it doesnt find
anything. Also, the orginal Koza experiment with this function used 50
points as opposed to the 20 points used by default in ecj. Should changing
it to 50 matter a lot? I would think it would make it a harder problem since
it has more things to match. In any case, I have tried with 50 as well and
it doesnt make a difference. 

I saw in a different post that Sean noted that Koza's experiment have been
hard to replicate, so I wonder whats the state with this one?

Thanks a lot,
Adelein