UnitDropout

class torch_brain.transforms.UnitDropout(field='spikes', reset_index=True, *args, **kwargs)

Bases: object

Augmentation that randomly drops units from the sample. By default, the number of units to keep is sampled from a triangular distribution defined in TriangleDistribution.

This transform assumes that the data has a units object. It works for both IrregularTimeSeries and RegularTimeSeries. For the former, it will drop spikes from the units that are not kept. For the latter, it will drop the corresponding columns from the data.

Parameters:

• field (str, optional) – Field to apply the dropout. Defaults to “spikes”.

• *args – Arguments to pass to the TriangleDistribution constructor.

• **kwargs – Arguments to pass to the TriangleDistribution constructor.