1Blurb:: Stopping criterion based on objective function or statistics convergence 2 3Description:: 4The \c convergence_tolerance specification provides a real value for 5controlling the termination of iteration. 6 7For optimization, it is most commonly a <b>relative convergence 8tolerance</b> for the objective function; i.e., if the change in the 9objective function between successive iterations divided by the 10previous objective function is less than the amount specified by 11convergence_tolerance, then this convergence criterion is satisfied on 12the current iteration. 13 14Therefore, permissible values are between 0 and 1, non-inclusive. 15 16<b> Behavior Varies by Package/Library </b> 17 18This control is used with most optimization and least squares 19iterators (DOT, CONMIN, NLPQLP, NPSOL, NLSSOL, OPT++, and SCOLIB). 20Most other Dakota methods (such as DACE or parameter studies) do not use 21this control, but some adaptive methods, such as adaptive UQ, do. 22 23Since no progress may be made on one iteration followed by significant 24progress on a subsequent iteration, some libraries require that the 25convergence tolerance be satisfied on two or more consecutive 26iterations prior to termination of iteration. 27 28Notes on each library: 29\li DOT: relative tolerance that must be satisfied for two consecutive iterations 30\li NL2SOL: See \ref method-nl2sol 31\li NLPQLP: used as Lagrangian gradient norm tolerance (ACC), not as a relative convergence tolerance 32\li NPSOL: used as a line search tolerance, not as a relative convergence tolerance 33 34Topics:: method_independent_controls 35Examples:: 36Theory:: 37Faq:: 38See_Also:: 39