I'm studying a problem where I have a potential V(x,y), a function of two variables, and I know that it is globally bounded from below and has three local minima. What I would like to do is identify the basins of attraction for each of the three minima. If I use a naive, crude, steepest descent algorithm, not the one in library, but a true:
xn = xn-1 - dt grad V(xn-1) I get what I'm after, though it takes more iterations than it should. When I try to use the algorithms in the library, my basins are inconsistent. I think this has to do with the adaptive selection of the step size, but I'm not entirely sure on how to control it. Is there a way to set a max step size in the algorithms? -gideon
