Thank you. In my application, it's important that I get to the same local minimum as if I hadn't stopped the minimizer. I believe saving x only won't do the trick.
I've been trying to stop/resume the minimizer by saving the object gsl_multimin_fminimizer, including its "void *state" pointer. Then, once I resume, I load up all of the relevant fields. However, I think the following command fails (which I found in fminimizer.c), probably because of something wrong while saving/loading s->state. (s->type->set) (s->state, s->f, s->x, &(s->size), step_size); where s is the gsl_multimin_fminimizer object. Thanks again. On Tue, Aug 6, 2019 at 9:56 PM Patrick Alken <[email protected]> wrote: > Can you just save the current position vector (state->x) and then start > from there the next time? > > On 8/6/19 5:06 PM, Pedro Gardete wrote: > > Hi, > > > > I'm running a minimization that may need to be interrupted some times, > due > > to computational constraints. > > > > Is there a way I can save the "gsl_multimin_minimizer" variable, so I can > > resume optimization? The tricky part is that this struct has a void > pointer > > with unknown size... I assume the remaining fields of the struct simply > > depend on the chosen minimization method. > > > > Thank you! > > Pedro > > >
