> I'm guessing that lbfgs is giving up early because the function is so
> flat. In the description in the manual lbfgs says that it uses a "low
> rank approximation to the Hessian matrix", perhaps this approximation
> breaks down with such a flat region?
Could be. I hope Robert will comment on this further.
> It is possible to create foreign function interfaces for C++ code, but
> this is dependent on which lisp you use for compiling maxima. I think
> there is no single FFI which works on all lisps supported by maxima.
Hmm, that sounds like a problem.
Kostas