abopt (ABstract OPTimizer) - optimization of generic numerical models
Geek Repo:Geek Repo
Github PK Tool:Github PK Tool
rainwoodman opened this issue 7 years ago · comments
The tol argument of LBFGS is relative, but GradientDescent is absolute. Does it make more sense to use absolute tolerance -- for we have been using convergence condition based on drop of chi square < 1, not relative drop of chi square.
tol
@aslanyan