tapas.attacks.set_classifiers.SetFeature

class tapas.attacks.set_classifiers.SetFeature

Bases: abc.ABC

Represents a set of features that can be extracted from a dataset, as a np.array vector. This is a callable that can optionally have some configuration parameters passed via __init__.

By default, this is memoryless and static, but it can be extended to be trainable from datasets.

__init__()

Methods

__init__()

extract(datasets)

Extract features from each dataset in a list.

Attributes

label

abstract extract(datasets: list[Dataset]) np.array

Extract features from each dataset in a list.

Parameters

datasets (list of Datasets.) – Datasets to extract features from.

Returns

features – Array of size len(datasets) x k, where the number of features k can be estimated by self.size(dataset.description). Each row is a dataset, and each column a different feature.

Return type

np.array