Package de.labathome.trustconstr.interfaces

Interface GlobalStoppingCriteria

public interface GlobalStoppingCriteria

Termination predicate consulted by the trust-region SQP and interior-point loops once per outer iteration. Implementations decide whether the algorithm has converged based on the current optimality and constraint- violation measures, the trust-region radius, the merit-function penalty, and IP-specific quantities (barrier parameter, inner-loop tolerance).

Method Summary

Modifier and Type	Method	Description
boolean	shouldStop(State state, Matrix x, boolean lastIterationFailed, double optimality, double constrViolation, double trustRadius, double penalty, CGInfo cgInfo, double barrierParameter, double tolerance)

Method Details

shouldStop

boolean shouldStop(State state,
 Matrix x,
 boolean lastIterationFailed,
 double optimality,
 double constrViolation,
 double trustRadius,
 double penalty,
 CGInfo cgInfo,
 double barrierParameter,
 double tolerance)

Parameters:

state - current outer-iteration state

x - current iterate

lastIterationFailed - whether the previous trial step was rejected

optimality - KKT optimality measure (infinity-norm of Lagrangian gradient)

constrViolation - infinity-norm of the constraint residual

trustRadius - current trust-region radius

penalty - current merit-function penalty

cgInfo - info from the projected-CG inner solve

barrierParameter - IP-only: current log-barrier coefficient

tolerance - IP-only: current barrier-subproblem tolerance

Returns:

true iff the outer loop should terminate