ShallowNetEncoder

class selfeeg.models.encoders.ShallowNetEncoder(Chans, F: int = 40, K1: int = 25, Pool: int = 75, p: float = 0.2, seed: int = None)[source]

Pytorch implementation of the ShallowNet Encoder.

See ShallowNet for some references. The expected input is a 3D tensor with size (Batch x Channels x Samples).

Parameters:

  • Chans (int) – The number of EEG channels.

  • F (int, optional) –

    The number of output filters in the temporal convolution layer.

    Default = 40

  • K1 (int, optional) –

    The length of the temporal convolutional layer.

    Default = 25

  • Pool (int, optional) –

    The temporal pooling kernel size.

    Default = 75

  • p (float, optional) –

    Dropout probability. Must be in [0,1)

    Default= 0.2

  • seed (int, optional) –

    A custom seed for model initialization. It must be a nonnegative number. If None is passed, no custom seed will be set

    Default = None

Note

In this implementation, the number of channels is an argument. However, in the original paper authors preprocess EEG data by selecting a subset of only 21 channels. Since the net is very minimalistic, please follow the authors’ notes.

Example

>>> import selfeeg.models
>>> import torch
>>> x = torch.randn(4,8,512)
>>> mdl = models.ShallowNetEncoder(8)
>>> out = mdl(x)
>>> print(out.shape) # shoud return torch.Size([4, 224])
>>> print(torch.isnan(out).sum()) # shoud return 0