search

SolverStoppingJulia / Stopping.jl

A framework to implement iterative algorithms
Other
12 stars 3 forks source link

`NLPStopping`: handle the unbounded below problem as a success in `optimality_check` for unconstrained problems #48

Open tmigot opened 2 years ago

tmigot commented 2 years ago

nlp_admissible_functions.jl#L12

Maybe add a keyword with an unbounded threshold.

tmigot commented 2 years ago

For constrained problems, we could replace the part "lagrangian gradient < eps" by 0 when the objective function is -Inf (or less than a threshold).

tmigot commented 2 years ago

The same should follow the 1D line searches