approximation_models
Specification of an unordered ensemble of low-fidelity approximations
Specification
Alias: unordered_model_fidelities
Arguments: None
JSON argument key: None
Description
An unordered or “non-hierarchical” ensemble surrogate manages an unordered set of low-fidelity model approximations, each of which may include hyper-parameter resolution controls (in the case of a simulation model) or additional model recursions.
Any corresponding sequence specifications within methods (e.g.,
quadrature_order_sequence, sparse_grid_level_sequence,
expansion_order_sequence, etc. within stochastic expansion methods)
should be synchronized with the order in the model listing.
Internal to the ensemble surrogate model, subsets of the model ensemble may be active for any given evaluation, as dictated by the iterative algorithm in use.
Examples
model
id_model = 'NONHIERARCH'
ensemble_surrogate
truth_model_pointer = 'HF'
approximation_models = 'LF1' 'LF2'
model
id_model = 'LF1'
single
interface_pointer = 'LF1_DRIVER'
solution_level_cost = 1.0
model
id_model = 'LF2'
single
interface_pointer = 'LF2_DRIVER'
solution_level_cost = 2.4
model
id_model = 'HF'
single
interface_pointer = 'HF_DRIVER'
solution_level_cost = 256.0
{
"model": [
{
"ensemble_surrogate": {
"id_model": "NONHIERARCH",
"ensemble": {
"truth_model_pointer": {
"pointer": "HF",
"approximation_models": ["LF1", "LF2"]
}
}
}
},
{
"single": {
"id_model": "LF1",
"interface_pointer": "LF1_DRIVER",
"cost_model": {
"solution_level_cost": [1.0]
}
}
},
{
"single": {
"id_model": "LF2",
"interface_pointer": "LF2_DRIVER",
"cost_model": {
"solution_level_cost": [2.4]
}
}
},
{
"single": {
"id_model": "HF",
"interface_pointer": "HF_DRIVER",
"cost_model": {
"solution_level_cost": [256.0]
}
}
}
]
}

