Improving Powell's method?


Alain Bourass
02-19-2004, 06:46 PM
Hi,
I am trying to fit a curve with a 16 parameters form. And, in order to reduce error between the model and reality, I used nr c implementation of Powell's method. Thus, I got an error of 0.012070 which is still bad.
So can someone tell me, on which default values playing in order to decrease that error rate. By default values, I mean GOLD, GLIMIT, TINY...

Thanks for your help and sorry for my english