tapas.attacks.distances.LpDistance

class tapas.attacks.distances.LpDistance(p: float = 2, weights: Optional[numpy.array] = None)

L_p distance between two datasets (typically, tabular datasets). This 1-hot encodes categorical attributes.

__init__(p: float = 2, weights: Optional[numpy.array] = None)

Parameters

p (float) – Order of the distance (default 2, Euclidean distance). p must be a positive number.
weights (real-valued numpy array) – Weighting to apply to individual entries in the 1-hot encoded dataset. The distance between records x and y (of length k) is computed as (sum_i weights_i * abs(x_i - y_i)^p )^(1/p).
of (Use the weights to restrict this distance to a specific subset) –
(e.g. (variables) –
columns). (to exclude 1-hot encoded) –

Methods

__init__([p, weights])

Attributes

label