mo.mo_mm

mo.mo_mm

Functions

Name	Description
mo_mm_desirability_function	Calculates the negative combined desirability for a candidate point x. Can be used by the mo_mm_desirability_optimizer.
mo_mm_desirability_optimizer	Optimizes the multi-objective function to find the next best point.
mo_xy_desirability_plot	Generates a plot of the desirability landscape.

mo_mm_desirability_function

mo.mo_mm.mo_mm_desirability_function(
    x,
    models,
    X_base,
    J_base,
    d_base,
    phi_base,
    D_overall,
    mm_objective=True,
    verbose=False,
)

Calculates the negative combined desirability for a candidate point x. Can be used by the mo_mm_desirability_optimizer. For each objective, a model is used to predict the objective value at x. If mm_objective is True, the Morris-Mitchell improvement is also calculated and included as an additional objective. The combined desirability, which uses the predictions from the models and optionally the Morris-Mitchell improvement, is then computed using the provided DOverall object.

Parameters

Name	Type	Description	Default
x	np.ndarray	Candidate point (1D array).	required
models	list	List of trained models. One model per objective.	required
X_base	np.ndarray	Existing design points. Used for computing Morris-Mitchell improvement.	required
J_base	np.ndarray	Multiplicities of distances for X_base. Used for Morris-Mitchell improvement.	required
d_base	np.ndarray	Unique distances for X_base. Used for Morris-Mitchell improvement.	required
phi_base	float	Base Morris-Mitchell metric for X_base. Used for Morris-Mitchell improvement.	required
D_overall	DOverall	The overall desirability function. Must include desirability functions for each objective and optionally for Morris-Mitchell.	required
mm_objective	bool	Whether to include space-filling improvement as an objective. Defaults to True.	True
verbose	bool	Whether to print Morris-Mitchell improvement values. Defaults to False.	False

Returns

Name	Type	Description
	Tuple[float, List[float]]	Tuple[float, List[float]]: A tuple containing: - Negative geometric mean of desirabilities (for minimization). - List of individual objective values.

Examples

>>> from spotoptim.mo import mo_mm_desirability_function
>>> from spotdesirability import DOverall, DMax
>>> import numpy as np
>>> from spotoptim.function.mo import mo_conv2_max
>>> from sklearn.ensemble import RandomForestRegressor
>>> from spotoptim.sampling.mm import mmphi_intensive
>>> # X_base in the range [0,1]
>>> X_base = np.random.rand(500, 2)
>>> y = mo_conv2_max(X_base)
>>> models = []
>>> for i in range(y.shape[1]):
...     model = RandomForestRegressor(n_estimators=100, random_state=42)
...     model.fit(X_base, y[:, i])
...     models.append(model)
>>> # calculate base Morris-Mitchell stats
>>> phi_base, J_base, d_base = mmphi_intensive(X_base, q=2, p=2)
>>> d_funcs = []
>>> for i in range(y.shape[1]):
...     d_func = DMax(low=np.min(y[:, i]), high=np.max(y[:, i]))
...     d_funcs.append(d_func)
>>> D_overall = DOverall(*d_funcs)
>>> x_test = np.random.rand(2)  # Example test point
>>> neg_D, objectives = mo_mm_desirability_function(x_test, models, X_base, J_base, d_base, phi_base, D_overall, mm_objective=False)
>>> print(f"Negative Desirability: {neg_D}")
Negative Desirability: ...
>>> print(f"Objectives: {objectives}")
Objectives: ...

mo_mm_desirability_optimizer

mo.mo_mm.mo_mm_desirability_optimizer(
    X_base,
    models,
    bounds,
    obj_func,
    **kwargs,
)

Optimizes the multi-objective function to find the next best point. Returns the best point, its desirability, and the history of objective values.

Parameters

Name	Type	Description	Default
X_base	np.ndarray	Existing design points.	required
models	list	List of trained surrogate models for each objective.	required
bounds	list	Bounds for each dimension.	required
obj_func	callable	Objective function to compute desirability and objectives.	required
**kwargs	Any	Additional arguments for the objective function.	{}

Returns

Name	Type	Description
	Tuple[np.ndarray, float, np.ndarray]	Tuple[np.ndarray, float, np.ndarray]: A tuple containing: - Best point (np.ndarray) - Best desirability (float) - History of objective values (np.ndarray)

mo_xy_desirability_plot

mo.mo_mm.mo_xy_desirability_plot(
    models,
    X_base,
    J_base,
    d_base,
    phi_base,
    D_overall,
    bounds=None,
    mm_objective=True,
    resolution=50,
    feature_names=None,
    **kwargs,
)

Generates a plot of the desirability landscape. Plots the 2-dim X values as points in the plane and colors them according to their desirability values. For each pair of inputs, x_i and x_j (with i < j), one plot is generated.

Parameters

Name	Type	Description	Default
models	list	List of trained models (one per objective).	required
X_base	np.ndarray	Existing design points.	required
J_base	np.ndarray	Multiplicities of distances for X_base.	required
d_base	np.ndarray	Unique distances for X_base.	required
phi_base	float	Base Morris-Mitchell metric.	required
D_overall	DOverall	The overall desirability function.	required
bounds	list	List of tuples (min, max) for each dimension. If None, derived from X_base.	None
mm_objective	bool	Whether to include space-filling improvement. Defaults to True.	True
resolution	int	Grid resolution for the plot. Defaults to 50.	50
feature_names	list	List of names for the input variables. Defaults to None.	None
**kwargs	Any	Additional arguments for plt.subplots (e.g., figsize).	{}

Returns

Name	Type	Description
	None	None

Examples

>>> from spotoptim.mo.mo_mm import mo_xy_desirability_plot
>>> import numpy as np
>>> from spotoptim.function.mo import mo_conv2_max
>>> from sklearn.ensemble import RandomForestRegressor
>>> from spotoptim.sampling.mm import mmphi_intensive
>>> # X_base in the range [0,1]
>>> X_base = np.random.rand(500, 2)
>>> y = mo_conv2_max(X_base)
>>> models = []
>>> for i in range(y.shape[1]):
...     model = RandomForestRegressor(n_estimators=100, random_state=42)
...     model.fit(X_base, y[:, i])
...     models.append(model)
>>> # calculate base Morris-Mitchell stats
>>> phi_base, J_base, d_base = mmphi_intensive(X_base, q=2, p=2)
>>> d_funcs = []
>>> for i in range(y.shape[1]):
...     d_func = DMax(low=np.min(y[:, i]), high=np.max(y[:, i]))
...     d_funcs.append(d_func)
>>> D_overall = DOverall(*d_funcs)
>>> mo_xy_desirability_plot(models, X_base, J_base, d_base, phi_base, D_overall)