belfem::opt Namespace Reference

Classes
class	Objective
	an object that defines a scalar objective function to be optimized. More...
class	Optimizer
	driver that optimizes an Objective using a chosen Algorithm. More...

Enumerations
enum class	Algorithm {   BOBYQA = 0 , COBYLA = 1 , NELDERMEAD = 2 , SBPLX = 3 ,   PRAXIS = 4 , MMA = 5 , SLSQP = 6 , LBFGS = 7 }
	optimization algorithm selector. More...
enum class	Status {   SUCCESS = 0 , STOPVAL_REACHED = 1 , FTOL_REACHED = 2 , XTOL_REACHED = 3 ,   MAXEVAL_REACHED = 4 , MAXTIME_REACHED = 5 , FAILURE = 6 , INVALID_ARGS = 7 ,   OUT_OF_MEMORY = 8 , ROUNDOFF_LIMITED = 9 , FORCED_STOP = 10 }
	outcome of an optimization run. More...

Functions
string	error_message (const Status aStatus)
	human-readable explanation of an optimizer status, suitable for a BELFEM_ERROR message when is_usable() is false.
bool	is_success (const Status aStatus)
	true if the status corresponds to a converged / usable result
bool	is_usable (const Status aStatus)
	true if the returned design point is usable, even when the solver did not terminate on a clean convergence criterion.

Enumeration Type Documentation

Algorithm

enum class belfem::opt::Algorithm

strong

optimization algorithm selector.

The names are library-agnostic; the mapping to the concrete nlopt algorithm lives in cl_Optimizer.cpp so that no nlopt header leaks into the public interface.

Naming follows nlopt's own convention where L = local, N = no derivatives, D = uses derivatives.

Enumerator
BOBYQA
COBYLA
NELDERMEAD
SBPLX
PRAXIS
MMA
SLSQP
LBFGS

Status

enum class belfem::opt::Status

strong

outcome of an optimization run.

The success codes mirror the termination reasons reported by the underlying library so the caller can tell why the solver stopped without depending on nlopt headers.

Enumerator
SUCCESS
STOPVAL_REACHED
FTOL_REACHED
XTOL_REACHED
MAXEVAL_REACHED
MAXTIME_REACHED
FAILURE
INVALID_ARGS
OUT_OF_MEMORY
ROUNDOFF_LIMITED
FORCED_STOP

Function Documentation

error_message()

string belfem::opt::error_message

(

const Status

aStatus

)

human-readable explanation of an optimizer status, suitable for a BELFEM_ERROR message when is_usable() is false.

Defined in cl_Optimizer.cpp. Analogous to MUMPS::error_message().

is_success()

bool belfem::opt::is_success ( const Status aStatus )

inline

true if the status corresponds to a converged / usable result

is_usable()

bool belfem::opt::is_usable ( const Status aStatus )

inline

true if the returned design point is usable, even when the solver did not terminate on a clean convergence criterion.

This is the MUMPS-style error/warning distinction: ROUNDOFF_LIMITED is the routine "converged as far as floating point allows" exit of the Powell-type algorithms ( BOBYQA/COBYLA ) under tight tolerances, and FORCED_STOP returns the best point found before the stop. Guard fatal aborts with this; use is_success() to detect a clean convergence.