ITMO_FS.filters.univariate.VDM

class ITMO_FS.filters.univariate.VDM(weighted=True)

Creates Value Difference Metric builder http://aura.abdn.ac.uk/bitstream/handle/2164/10951/payne_ecai_98.pdf?sequence=1 https://www.jair.org/index.php/jair/article/view/10182

Parameters:weighted (bool) – If weighted = False, modified version of metric which omits the weights is used

Notes

For more details see papers about Improved Heterogeneous Distance Functions and Implicit Future Selection with the VDM.

Examples

>>> x = np.array([[0, 0, 0, 0],
...               [1, 0, 1, 1],
...               [1, 0, 0, 2]])
>>> y = np.array([0,
...               1,
...               1])
>>> vdm = VDM()
>>> vdm.run(x, y)
array([[0.         4.35355339 4.        ]
       [4.5        0.         0.5       ]
       [4.         0.35355339 0.        ]])

__init__(weighted=True)

Initialize self. See help(type(self)) for accurate signature.

run(x, y)

Generates metric for the data Complexity: O(n_features * n_samples^3) worst case, should be faster on a real data

x: array-like, shape (n_features, n_samples)

Input samples’ parameters. Parameters among every class must be sequential integers.

y: array-like, shape (n_samples)

Input samples’ class labels. Class labels must be sequential integers.

result:

numpy.ndarray, shape=(n_samples, n_samples), dtype=np.double with selected version of metrics

feature_scores = {} def run(self, x, y, weighted=True):