braindecode.models.USleep¶
- class braindecode.models.USleep(in_chans=2, sfreq=128, depth=12, n_time_filters=5, complexity_factor=1.67, with_skip_connection=True, n_classes=5, input_size_s=30, time_conv_size_s=0.0703125, ensure_odd_conv_size=False, apply_softmax=False)¶
Sleep staging architecture from Perslev et al 2021.
U-Net (autoencoder with skip connections) feature-extractor for sleep staging described in [1].
- For the encoder (‘down’):
– the temporal dimension shrinks (via maxpooling in the time-domain) – the spatial dimension expands (via more conv1d filters in the
time-domain)
- For the decoder (‘up’):
– the temporal dimension expands (via upsampling in the time-domain) – the spatial dimension shrinks (via fewer conv1d filters in the
time-domain)
Both do so at exponential rates.
- Parameters
- in_chansint
Number of EEG or EOG channels. Set to 2 in [1] (1 EEG, 1 EOG).
- sfreqfloat
EEG sampling frequency. Set to 128 in [1].
- depthint
Number of conv blocks in encoding layer (number of 2x2 max pools) Note: each block halve the spatial dimensions of the features.
- n_time_filtersint
Initial number of convolutional filters. Set to 5 in [1].
- complexity_factorfloat
Multiplicative factor for number of channels at each layer of the U-Net. Set to 2 in [1].
- with_skip_connectionbool
If True, use skip connections in decoder blocks.
- n_classesint
Number of classes. Set to 5.
- input_size_sfloat
Size of the input, in seconds. Set to 30 in [1].
- time_conv_size_sfloat
Size of the temporal convolution kernel, in seconds. Set to 9 / 128 in [1].
- ensure_odd_conv_sizebool
If True and the size of the convolutional kernel is an even number, one will be added to it to ensure it is odd, so that the decoder blocks can work. This can ne useful when using different sampling rates from 128 or 100 Hz.
- apply_softmaxbool
If True, apply softmax on output (e.g. when using nn.NLLLoss). Use False if using nn.CrossEntropyLoss.
References
- 1(1,2,3,4,5,6,7)
Perslev M, Darkner S, Kempfner L, Nikolic M, Jennum PJ, Igel C. U-Sleep: resilient high-frequency sleep staging. npj Digit. Med. 4, 72 (2021). https://github.com/perslev/U-Time/blob/master/utime/models/usleep.py
Methods
- forward(x)¶
If input x has shape (B, S, C, T), return y_pred of shape (B, n_classes, S). If input x has shape (B, C, T), return y_pred of shape (B, n_classes).
