variable_tolerance
Step length-based stopping criteria for derivative-free optimizers
Specification
Alias: None
Arguments: None
JSON argument key: None
Default: 1e-06
Description
The variable_tolerance keyword defines the minimum step length allowed
by the optimizer and is used to determine convergence. It is
applicable to asynch_pattern_search, coliny_cobyla,
coliny_pattern_search, coliny_solis_wets, and
mesh_adaptive_search.
Default Behavior
The default value varies according to method as follows:
asynch_pattern_search: 1.0e-2coliny_cobyla: 1.0e-4coliny_pattern_search: 1.0e-5coliny_solis_wets: 1.0e-6mesh_adaptive_search: 1.0e-6
Usage Tips
It is recommended that variable_tolerance be set to a value for which
changes of that scale in parameter values cause negligible changes in
the objective function.
Examples
Three example method input blocks appear below.
For asynch_pattern_search:
method
asynch_pattern_search
contraction_factor = 0.25
variable_tolerance = 1.e-4
solution_target = 1.e-6
max_function_evaluations 500
constraint_tolerance 1.e-6
{
"method": [
{
"asynch_pattern_search": {
"contraction_factor": 0.25,
"variable_tolerance": 1e-4,
"solution_target": 1e-6,
"max_function_evaluations": 500,
"constraint_tolerance": 1e-6
}
}
]
}
For coliny_pattern_search:
method
coliny_pattern_search
initial_delta = .2
variable_tolerance = 1.e-4
max_iterations 100
solution_accuracy = 1.e-6
seed = 1234
max_function_evaluations = 1000
{
"method": [
{
"coliny_pattern_search": {
"initial_delta": 0.2,
"variable_tolerance": 1e-4,
"max_iterations": 100,
"solution_accuracy": 1e-6,
"seed": 1234,
"max_function_evaluations": 1000
}
}
]
}
For mesh_adaptive_search
method
mesh_adaptive_search
variable_tolerance = 0.01
seed = 1234
{
"method": [
{
"mesh_adaptive_search": {
"variable_tolerance": 0.01,
"seed": 1234
}
}
]
}

