Is it possible to use Newton() methods with limiting some variables? Like FMinbox?

You might have better luck transforming your variables, as done here: Optim.jl . They work with the log variance which can take on any value. For ρ you could use tanh and atanh to go back and forth between (-1, 1) and (-inf, inf)

Optim.jl: bracketing with Newton()

Specific Domains Optimization (Mathematical)

Tamas_Papp October 28, 2019, 9:47am 8

It is hard to say more about this without an MWE (eg do you provide derivatives manually or use AD, what is rho [stepsize correction?], …).