Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: help in genetic algorithm
Posted:
Apr 8, 2013 12:55 PM


On 4/8/2013 12:29 PM, mado wrote: > Alan_Weiss <aweiss@mathworks.com> wrote in message > <kjudjf$r7m$1@newscl01ah.mathworks.com>... >> On 4/8/2013 7:02 AM, mado wrote: >> > i want to ask about genetic algorithm , i am using it to tune pid >> > parameters but it gives different p, i d values in every running >> and > it?s not satisfying , need your help >> > i don't know is this problem of ga parameters or it's the objective >> > function >> >> The genetic algorithm is stochastic, it often gives different results >> every time. >> >> Allow me to suggest that you use patternsearch instead of ga as your >> solver. It is usually faster, more robust, and easier to tune. The >> only down side is you need to supply initial points. If you have >> finite upper and lower bounds, you can try >> >> x0 = lb + rand(size(lb)).*(ub  lb); >> >> Good luck, >> > > I need to try ga , i increased generations but don't affect my problem > is the values is small i need it to be bigger and it should be to > minimize error being small that mean there is a big error between > reference and output value
The most important thing to modify is the initial range. http://www.mathworks.com/help/gads/populationdiversity.html
The next most important thing is the population size.
But the best thing to do is chuck ga and use patternsearch.
Alan Weiss MATLAB mathematical toolbox documentation



