Hi, thanks for sharing a great (and necessary) piece of code. Ive got a few questions regarding LBFGSB.
Im using the v2 version, works fine on my case but i notice that my Solver-Statex_delta and f_delta become considerably small leading to nan function values, Any suggestions on how I can trigger an early stop? or maybe even a restart?
EDIT:
How do I instantiate a cppoptlib::solver::State struct such that i could override the DefaultStoppingSolverState()?
Hi, thanks for sharing a great (and necessary) piece of code. Ive got a few questions regarding LBFGSB. Im using the v2 version, works fine on my case but i notice that my
Solver-State
x_delta
andf_delta
become considerably small leading tonan
function values, Any suggestions on how I can trigger an early stop? or maybe even a restart?EDIT: How do I instantiate a
cppoptlib::solver::State
struct such that i could override theDefaultStoppingSolverState()
?Thanks