tricands.tricands

tricands.tricands

Functions

Name	Description
tricands	Generate Triangulation Candidates for Bayesian Optimization.
tricands_fringe	Generate fringe candidates outside the convex hull.
tricands_interior	Generate interior candidates using Delaunay triangulation.

tricands

tricands.tricands.tricands(
    X,
    p=0.5,
    fringe=True,
    nmax=None,
    best=None,
    ordering=None,
    vis=False,
    imgname='tricands.pdf',
    lower=0,
    upper=1,
)

Generate Triangulation Candidates for Bayesian Optimization. Assumes a bounding box of [lower, upper]^m.

Parameters

Name	Type	Description	Default
X	np.ndarray	Design matrix of shape (n_samples, n_features). Each row gives a design point and each column a feature.	required
p	float	Distance to the boundary for fringe candidates (0 = on hull, 1 = on boundary). Defaults to 0.5.	0.5
fringe	bool	Whether to include fringe points to allow exploration outside the convex hull. Defaults to True.	True
nmax	int	Maximum size of candidate set. If output exceeds this, strategic subsetting is employed. Defaults to 100 * n_features.	None
best	int	Index of the best (lowest) currently observed point. Used for strategic subsetting in Bayesian optimization. Defaults to None.	None
ordering	np.ndarray	Order of closeness of rows of X to a contour level. Used for contour location subsetting. Defaults to None.	None
vis	bool	Whether to visualize the triangulation. Only applicable to 2D designs. Defaults to False.	False
imgname	str	File name for saved plot if vis=True. Defaults to ‘tricands.pdf’.	'tricands.pdf'
lower	float	Lower bound of bounding box for all dimensions. Defaults to 0.	0
upper	float	Upper bound of bounding box for all dimensions. Defaults to 1.	1

Returns

Name	Type	Description
	np.ndarray	np.ndarray: Array of candidate points, shape (n_candidates, n_features).

Raises

Name	Type	Description
	Exception	If visualization is requested for non-2D data.
	Exception	If number of points is less than n_features + 1.
	Exception	If both ‘best’ and ‘ordering’ are provided.
	Exception	If X contains values outside [lower, upper].

Examples

>>> import numpy as np
>>> from spotoptim.tricands import tricands
>>> X = np.array([[0.1, 0.1], [0.9, 0.1], [0.5, 0.9], [0.2, 0.5]])
>>> candidates = tricands(X, fringe=True, p=0.5)
>>> print(candidates.shape)
(7, 2)

tricands_fringe

tricands.tricands.tricands_fringe(X, p=0.5, lower=0, upper=1)

Generate fringe candidates outside the convex hull.

Subroutine used by tricands wrapper. Assumes a bounding box of [lower, upper]^m.

Parameters

Name	Type	Description	Default
X	np.ndarray	Input design matrix of shape (n_samples, n_features).	required
p	float	Distance to the boundary (0 = on hull, 1 = on boundary). Defaults to 0.5.	0.5
lower	float	Lower bound of bounding box for all dimensions. Defaults to 0.	0
upper	float	Upper bound of bounding box for all dimensions. Defaults to 1.	1

Returns

Name	Type	Description
dict	dict	A dictionary containing: - ‘XF’ (np.ndarray): Fringe candidate points. - ‘XB’ (np.ndarray): Boundary points (means of external facets). - ‘qhull’ (scipy.spatial.ConvexHull): The computed convex hull object.

Raises

Name	Type	Description
	Exception	If the number of points is less than n_features + 1.

tricands_interior

tricands.tricands.tricands_interior(X)

Generate interior candidates using Delaunay triangulation.

Subroutine used by tricands wrapper.

Parameters

Name	Type	Description	Default
X	np.ndarray	Input design matrix of shape (n_samples, n_features).	required

Returns

Name	Type	Description
dict	dict	A dictionary containing: - ‘cand’ (np.ndarray): Candidate points (midpoints of triangles). - ‘tri’ (np.ndarray): Simplicies of the Delaunay triangulation.

Raises

Name	Type	Description
	Exception	If the number of points is less than n_features + 1.