I have one question - do you plan to support an option to use a symbolic (so, not numerically approximated) Hessian? I have some small-scale problem which requires high accuracy and therefore the symbolic Hessian suits better than numerical one.
L-BFGS is a quasi-Newton method, meaning that it does not require an exact Hessian matrix. If you have access to the Hessian matrix, then you can directly apply the Newton's method.
Hello! It is an awesome library, thank you!
I have one question - do you plan to support an option to use a symbolic (so, not numerically approximated) Hessian? I have some small-scale problem which requires high accuracy and therefore the symbolic Hessian suits better than numerical one.