bluemira.optimisation._nlopt.functions

Attributes

_FloatOrArrayT

Classes

ConstraintType

Enumeration of constraint types.

_NloptFunction

Base class for NLOpt objective/constraint functions.

ObjectiveFunction

Holds an objective function for an NLOpt optimiser.

Constraint

Holder for NLOpt constraint functions.

Module Contents

bluemira.optimisation._nlopt.functions._FloatOrArrayT
class bluemira.optimisation._nlopt.functions.ConstraintType(*args, **kwds)

Bases: enum.Enum

Inheritance diagram of bluemira.optimisation._nlopt.functions.ConstraintType

Enumeration of constraint types.

EQUALITY
INEQUALITY
class bluemira.optimisation._nlopt.functions._NloptFunction(f: collections.abc.Callable[[numpy.ndarray], _FloatOrArrayT], bounds: tuple[_FloatOrArrayT, _FloatOrArrayT])

Base class for NLOpt objective/constraint functions.

Implements the ability to calculate a numerical gradient, if an analytical one is not provided.

Parameters:

  • f (collections.abc.Callable[[numpy.ndarray], _FloatOrArrayT])

  • bounds (tuple[_FloatOrArrayT, _FloatOrArrayT])

f
f0: float | numpy.ndarray = 0
bounds
_approx_derivative(x: numpy.ndarray) → numpy.ndarray
Returns:

The approximated derivative of the function at x.

Parameters:

x (numpy.ndarray)

Return type:

numpy.ndarray

set_approx_derivative_lower_bound(lower_bound: numpy.ndarray) → None

Set the lower bounds for use in derivative approximation.

Parameters:

lower_bound (numpy.ndarray)

Return type:

None

set_approx_derivative_upper_bound(upper_bound: numpy.ndarray) → None

Set the upper bounds for use in derivative approximation.

Parameters:

upper_bound (numpy.ndarray)

Return type:

None

class bluemira.optimisation._nlopt.functions.ObjectiveFunction(f: bluemira.optimisation.typed.ObjectiveCallable, df: bluemira.optimisation.typed.OptimiserCallable | None, n_variables: int, bounds: tuple[_FloatOrArrayT, _FloatOrArrayT] = (-np.inf, np.inf))

Bases: _NloptFunction

Inheritance diagram of bluemira.optimisation._nlopt.functions.ObjectiveFunction

Holds an objective function for an NLOpt optimiser.

Adapts the given objective function, and optional derivative, to a form understood by NLOpt.

If no optimiser derivative is given, and the algorithm is gradient based, a numerical approximation of the gradient is calculated.

Parameters:
f: bluemira.optimisation.typed.ObjectiveCallable
f0: float
df
history: list[tuple[numpy.ndarray, float]] = []
prev_iter
call(x: numpy.ndarray, grad: numpy.ndarray) → float

Execute the NLOpt objective function.

Returns:

The value of the objective function at x.

Parameters:

  • x (numpy.ndarray)

  • grad (numpy.ndarray)

Return type:

float

call_with_history(x: numpy.ndarray, grad: numpy.ndarray) → float

Execute the NLOpt objective function, recording the iteration history.

Returns:

The value of the objective function at x, with the iteration history recorded.

Parameters:

  • x (numpy.ndarray)

  • grad (numpy.ndarray)

Return type:

float

_call_inner(x: numpy.ndarray, grad: numpy.ndarray) → float

Execute the objective function in the form required by NLOpt.

Returns:

The value of the objective function at x.

Parameters:

  • x (numpy.ndarray)

  • grad (numpy.ndarray)

Return type:

float

_store_x(x: numpy.ndarray) → None

Store x in self.prev_iter.

Parameters:

x (numpy.ndarray)

Return type:

None

class bluemira.optimisation._nlopt.functions.Constraint(constraint_type: ConstraintType, f: bluemira.optimisation.typed.OptimiserCallable, tolerance: numpy.ndarray, df: bluemira.optimisation.typed.OptimiserCallable | None = None, bounds: tuple[_FloatOrArrayT, _FloatOrArrayT] = (-np.inf, np.inf))

Bases: _NloptFunction

Inheritance diagram of bluemira.optimisation._nlopt.functions.Constraint

Holder for NLOpt constraint functions.

Parameters:
f: bluemira.optimisation.typed.OptimiserCallable
f0: numpy.ndarray
constraint_type
tolerance
df
history: list[tuple[numpy.ndarray, Ellipsis]] = []
call(result: numpy.ndarray, x: numpy.ndarray, grad: numpy.ndarray) → None

Execute the constraint function in the form required by NLOpt.

https://nlopt.readthedocs.io/en/latest/NLopt_Python_Reference/#vector-valued-constraints

Parameters:

  • result (numpy.ndarray)

  • x (numpy.ndarray)

  • grad (numpy.ndarray)

Return type:

None

call_with_history(result: numpy.ndarray, x: numpy.ndarray, grad: numpy.ndarray) → None
Parameters:

  • result (numpy.ndarray)

  • x (numpy.ndarray)

  • grad (numpy.ndarray)

Return type:

None