Source code for torcheeg.models.cnn.fbcnet
import torch
import torch.nn as nn
class Conv2dWithConstraint(nn.Conv2d):
def __init__(self, *args, weight_norm=True, max_norm=1, **kwargs):
self.max_norm = max_norm
self.weight_norm = weight_norm
super(Conv2dWithConstraint, self).__init__(*args, **kwargs)
def forward(self, x):
if self.weight_norm:
self.weight.data = torch.renorm(self.weight.data,
p=2,
dim=0,
maxnorm=self.max_norm)
return super(Conv2dWithConstraint, self).forward(x)
class LinearWithConstraint(nn.Linear):
def __init__(self, *args, weight_norm=True, max_norm=1, **kwargs):
self.max_norm = max_norm
self.weight_norm = weight_norm
super(LinearWithConstraint, self).__init__(*args, **kwargs)
def forward(self, x):
if self.weight_norm:
self.weight.data = torch.renorm(self.weight.data,
p=2,
dim=0,
maxnorm=self.max_norm)
return super(LinearWithConstraint, self).forward(x)
class VarLayer(nn.Module):
def __init__(self, dim):
super(VarLayer, self).__init__()
self.dim = dim
def forward(self, x):
return x.var(dim=self.dim, keepdim=True)
class StdLayer(nn.Module):
def __init__(self, dim):
super(StdLayer, self).__init__()
self.dim = dim
def forward(self, x):
return x.std(dim=self.dim, keepdim=True)
class LogVarLayer(nn.Module):
def __init__(self, dim):
super(LogVarLayer, self).__init__()
self.dim = dim
def forward(self, x):
return torch.log(
torch.clamp(x.var(dim=self.dim, keepdim=True), 1e-6, 1e6))
class MeanLayer(nn.Module):
def __init__(self, dim):
super(MeanLayer, self).__init__()
self.dim = dim
def forward(self, x):
return x.mean(dim=self.dim, keepdim=True)
class MaxLayer(nn.Module):
def __init__(self, dim):
super(MaxLayer, self).__init__()
self.dim = dim
def forward(self, x):
ma, ima = x.max(dim=self.dim, keepdim=True)
return ma
class swish(nn.Module):
def __init__(self):
super(swish, self).__init__()
def forward(self, x):
return x * torch.sigmoid(x)
[docs]class FBCNet(nn.Module):
r'''
An Efficient Multi-view Convolutional Neural Network for Brain-Computer Interface. For more details, please refer to the following information.
- Paper: Mane R, Chew E, Chua K, et al. FBCNet: A multi-view convolutional neural network for brain-computer interface[J]. arXiv preprint arXiv:2104.01233, 2021.
- URL: https://arxiv.org/abs/2104.01233
- Related Project: https://github.com/ravikiran-mane/FBCNet
Below is a recommended suite for use in emotion recognition tasks:
.. code-block:: python
dataset = DEAPDataset(io_path=f'./deap',
root_path='./data_preprocessed_python',
chunk_size=512,
num_baseline=1,
baseline_chunk_size=512,
offline_transform=transforms.BandSignal(),
online_transform=transforms.ToTensor(),
label_transform=transforms.Compose([
transforms.Select('valence'),
transforms.Binary(5.0),
]))
model = FBCNet(num_classes=2,
num_electrodes=32,
chunk_size=512,
in_channels=4,
num_S=32)
Args:
num_electrodes (int): The number of electrodes. (default: :obj:`28`)
chunk_size (int): Number of data points included in each EEG chunk. (default: :obj:`1000`)
in_channels (int): The number of channels of the signal corresponding to each electrode. If the original signal is used as input, in_channels is set to 1; if the original signal is split into multiple sub-bands, in_channels is set to the number of bands. (default: :obj:`9`)
num_S (int): The number of spatial convolution block. (default: :obj:`32`)
num_classes (int): The number of classes to predict. (default: :obj:`2`)
temporal (str): The temporal layer used, with options including VarLayer, StdLayer, LogVarLayer, MeanLayer, and MaxLayer, used to compute statistics using different techniques in the temporal dimension. (default: :obj:`LogVarLayer`)
stride_factor (int): The stride factor. (default: :obj:`4`)
weight_norm (bool): Whether to use weight renormalization technique in Conv2dWithConstraint. (default: :obj:`True`)
'''
def __init__(self,
num_electrodes: int = 20,
chunk_size: int = 1000,
in_channels: int = 9,
num_S: int = 32,
num_classes: int = 2,
temporal: str = 'LogVarLayer',
stride_factor: int = 4,
weight_norm: bool = True):
super(FBCNet, self).__init__()
self.num_electrodes = num_electrodes
self.chunk_size = chunk_size
self.num_classes = num_classes
self.in_channels = in_channels
self.num_S = num_S
self.temporal = temporal
self.stride_factor = stride_factor
self.weight_norm = weight_norm
assert chunk_size % stride_factor == 0, f'chunk_size should be divisible by stride_factor, chunk_size={chunk_size},stride_factor={stride_factor} does not meet the requirements.'
self.scb = self.SCB(num_S,
num_electrodes,
self.in_channels,
weight_norm=weight_norm)
if temporal == 'VarLayer':
self.temporal_layer = VarLayer(dim=3)
elif temporal == 'StdLayer':
self.temporal_layer = StdLayer(dim=3)
elif temporal == 'LogVarLayer':
self.temporal_layer = LogVarLayer(dim=3)
elif temporal == 'MeanLayer':
self.temporal_layer = MeanLayer(dim=3)
elif temporal == 'MaxLayer':
self.temporal_layer = MaxLayer(dim=3)
else:
raise NotImplementedError
self.last_layer = self.last_block(self.num_S * self.in_channels *
self.stride_factor,
num_classes,
weight_norm=weight_norm)
def SCB(self, num_S, num_electrodes, in_channels, weight_norm=True):
return nn.Sequential(
Conv2dWithConstraint(in_channels,
num_S * in_channels, (num_electrodes, 1),
groups=in_channels,
max_norm=2,
weight_norm=weight_norm,
padding=0),
nn.BatchNorm2d(num_S * in_channels), swish())
def last_block(self, in_channels, out_channels, weight_norm=True):
return nn.Sequential(
LinearWithConstraint(in_channels,
out_channels,
max_norm=0.5,
weight_norm=weight_norm), nn.LogSoftmax(dim=1))
[docs] def forward(self, x: torch.Tensor) -> torch.Tensor:
r'''
Args:
x (torch.Tensor): EEG signal representation, the ideal input shape is :obj:`[n, in_channel, num_electrodes, chunk_size]`. Here, :obj:`n` corresponds to the batch size
Returns:
torch.Tensor[number of sample, number of classes]: the predicted probability that the samples belong to the classes.
'''
x = self.scb(x)
x = x.reshape([
*x.shape[0:2], self.stride_factor,
int(x.shape[3] / self.stride_factor)
])
x = self.temporal_layer(x)
x = torch.flatten(x, start_dim=1)
x = self.last_layer(x)
return x
def feature_dim(self):
return self.num_S * self.in_channels * self.stride_factor
