braindecode.visualization.random_target

braindecode.visualization.random_target(target, n_classes, generator=None)

Return labels uniformly sampled from {0, ..., n_classes-1} \ target.

For each entry of target pick a different class at random. Used in the label-randomization sanity check: query the trained model’s attribution method with the wrong target and check whether the resulting map differs from the correct-target map. Accepts a Python int, NumPy array, or torch tensor and returns the same kind of object on the same device.

Parameters:

• target (int, numpy.ndarray, or torch.Tensor) – True class index/indices.

• n_classes (int) – Total number of classes; must be at least 2.

• generator (numpy.random.Generator, optional) – Source of randomness. Defaults to numpy.random.default_rng().

Return type:

Same type as target (or int when target is a scalar).

Examples using braindecode.visualization.random_target

Interpretability of EEG Decoders

Interpretability of EEG Decoders