Hi, everybody I've a real problem with some of Numerical Recipes routines in f77. I'm trying to isolate the minimum of a quadratic multidimensional equation (i.e. (x+2)^2+(y+2)^2+3) which will the most simple case with the Conjugate Gradient Method. I use the routine mnbrak dbrent and dlinmin that you can easily find on the Numerical recipes home page. (If your are interested, I can send you the source files of "my" code). The problem is the dependancy of the solution ( i.e. the location of minimum ) to the INITIAL GUESS. If anybody have used those routines...

for f(x,y)= (x+2)^2+(y+2)^2+3) the minimum is f(-2,-2)=3

HA............ Good Job............

That is normal, to a certain extend, with non-linear optimization