cosapp.systems.metamodels

System tuned to support meta-models.

Classes

MetaSystem(name, data, mapping, ...)

Based System to insert meta-system inside CoSApp.

class cosapp.systems.metamodels.MetaSystem(name: str, data: str | ~pandas.core.frame.DataFrame, mapping: ~typing.Dict[str, str] | None = None, default_model: ~typing.Type[~cosapp.utils.surrogate_models.base.SurrogateModel] = <class 'cosapp.utils.surrogate_models.nearest_neighbor.NearestNeighbor'>, **kwargs)

Bases: System

Based System to insert meta-system inside CoSApp.

Parameters:

• data (str or pandas.DataFrame)

• mapping (Dict[str, str], optional)

• default_model (Type[SurrogateModel], optional) – Default surrogate model; default NearestNeighbor with RBF interpolator

• name (str, optional) – Name of the System; default ‘undefined’

models

Dictionary of surrogate model for each output variables

Type:

dict[str, SurrogateModel]

_trained

Are the surrogate models trained?

Type:

bool

_default_model

Default surrogate model to use

Type:

Type[SurrogateModel]

_input_names

List of port variable names (port.`variable`)

Type:

list of str

Examples

As for classical System, this class needs to be subclassed to be meaning full.

>>> class XPort(Port):
>>>     def setup(self):
>>>         self.add_variable('x')
>>>
>>> class OutPort(Port):
>>>     def setup(self):
>>>         self.add_variable('sin')
>>>         self.add_variable('cos')
>>>
>>> class MetaA(MetaSystem):
>>>     def setup(self):
>>>         self.add_input(XPort, 'in_')
>>>         self.add_output(OutPort, 'out', model=NearestNeighbor)

add_output(port_class: Type[Port], name: str, variables: Dict[str, Any] | None = None, model: Type[SurrogateModel] | Dict[str, SurrogateModel] | None = None) → Port

Add an output Port to the System.

This function cannot be called outside System.setup.

Parameters:

• port_class (type) – Class of the Port to create

• name (str) – Port name

• variables (Dict[str, Any], optional) – Dictionary of initial values (default: None)

• model (Type[Surrogate] or Dict[str, SurrogateModel], optional) – Surrogate default model type or mapping of surrogate model with output names; default NearestNeighbor.

Returns:

The created port

Return type:

Port

Examples

>>> class MyPort(Port):
>>>     def setup(self):
>>>         self.add_variable('x')
>>>         self.add_variable('y')
>>>
>>> class MySystem(System):
>>>     def setup(self):
>>>         self.add_output(MyPort, 'output_x')
>>>         self.add_output(MyPort, 'output_y', model=NearestNeighbor)
>>>         self.add_output(MyPort, 'output_z', model={'x': NearestNeighbor,
>>>                                                    'y': FloatKrigingSurrogate})

add_outward(definition: str | Dict[str, Any], value: Any = 1, unit: str = '', dtype: type | None = None, valid_range: Tuple[Any, Any] | None = None, invalid_comment: str = '', limits: Tuple[Any, Any] | None = None, out_of_limits_comment: str = '', desc: str = '', scope: Scope = Scope.PUBLIC, model: Type[SurrogateModel] | Dict[str, SurrogateModel] | None = None) → None

Add a outward variable to the System.

A outward variable is calculated by the System. But its value is not mandatory in any variables fluxes between this System and another one.

An unique outward variable can be defined by providing directly all arguments. And multiple outward variables can be defined by passing a dict of pair (str, Any) with an entry for each variable.

This function cannot be called outside setup().

Parameters:

• definition (str or Dict[str, Any]) – Name of the unique variable or a dictionary for multiple variables at once

• value (Any, optional) – Value of the variable if definition is a str; default 1

• unit (str, optional) – Variable unit; default empty string (i.e. dimensionless)

• dtype (type or iterable of types, optional) – Variable type; default None (i.e. type of initial value)

• valid_range (Tuple[Any, Any], optional) – Validity range of the variable; default None (i.e. all values are valid)

• invalid_comment (str, optional) – Comment to show in case the value is not valid; default ‘’

• limits (Tuple[Any, Any], optional) – Limits over which the use of the model is wrong; default valid_range

• out_of_limits_comment (str, optional) – Comment to show in case the value is not valid; default ‘’

• desc (str, optional) – Variable description; default ‘’

• scope (Scope {PRIVATE, PROTECTED, PUBLIC}, optional) – Variable visibility; default PUBLIC

Examples

To add an unique variable, arguments must be directly specified.

>>> system.add_outward('info', 2.)

To add multiple variables, a dictionary with one key per outward variable should be provided.

>>> system.add_inward({
>>>     'info1': 42.,
>>>     'info2': False
>>> })

compute() → None

Contains the customized System calculation.

models

rebuild_out_sequences()

split_in_sequences()

split_sequences_in_training_data(port_name, var_name)

update()

Perform complex tasks due to data change.

Some parameters may be a file used to initialize a complex object. Updating that parameter may imply the modification of that complex object. This should be done in this method.

Examples

>>> class TableModel(System):
>>>
>>>     def setup(self):
>>>         # Add a inwards to the file containing the table values
>>>         self.add_inward('table_file', '{myProject}/ressources/my_table.csv')
>>>         # Set a object able to interpolate the table
>>>         self.add_outward('table', Table(self.table_file))
>>>
>>>     def update(self):
>>>         # Update the table object as the source file may have been updated.
>>>         self.table = Table(self.table_file)

Notes

This method is called systematically after the setup() call. Otherwise the user is responsible to explicitly call it when needed.