1Blurb::
2Step length-based stopping criteria for derivative-free optimizers
3
4Description::
5The \c variable_tolerance keyword defines the minimum step length allowed
6by the optimizer and is used to determine convergence.  It is
7applicable to \c asynch_pattern_search, \c coliny_cobyla, \c
8coliny_pattern_search, \c coliny_solis_wets, and \c
9mesh_adaptive_search.
10
11<b> Default Behavior </b>
12
13The default value varies according to method as follows:
14
15\li \c asynch_pattern_search:  1.0e-2
16\li \c coliny_cobyla:  1.0e-4
17\li \c coliny_pattern_search:  1.0e-5
18\li \c coliny_solis_wets:  1.0e-6
19\li \c mesh_adaptive_search:  1.0e-6
20
21<b> Usage Tips </b>
22
23It is recommended that \c variable_tolerance be set to a value for which
24changes of that scale in parameter values cause negligible changes in
25the objective function.
26
27Topics::
28
29Examples::
30
31Three example method input blocks appear below.
32
33For \c asynch_pattern_search:
34\verbatim
35method
36  asynch_pattern_search
37    contraction_factor = 0.25
38    variable_tolerance = 1.e-4
39    solution_target = 1.e-6
40    max_function_evaluations 500
41    constraint_tolerance    1.e-6
42\endverbatim
43
44For \c coliny_pattern_search:
45\verbatim
46method
47  coliny_pattern_search
48    initial_delta = .2
49    variable_tolerance = 1.e-4
50    max_iterations 100
51    solution_accuracy = 1.e-6
52    seed = 1234
53    max_function_evaluations = 1000
54\endverbatim
55
56For \c mesh_adaptive_search
57\verbatim
58method
59  mesh_adaptive_search
60    variable_tolerance = 0.01
61    seed = 1234
62\endverbatim
63
64Theory::
65Faq::
66See_Also::
67