I have found that the default parameters work well about 99% of the time.
So, you really don't have to set any parameters unless your training is having a problem with the defaults.
I typically design 100 nets at a time. If training is slow (typically because of a large training set), I increase the MSEgoal and MinGrad to shorten training time
net.trainParam.goal = MSEgoal = 0.01*mean(var(target',1))
net.trainParam.min_grad = MinGrad = MSEgoal/100
When mse=MSEgoal, the net has successfully modeled 99% of the target variance. That is good enough for me.
For examples, search the NEWSGROUP and ANSWERS using
Hope this helps.
Thank you for formally accepting my answer
Greg
Best Answer