arjun_vit.py

from typing import Tuple

import torch
from einops import rearrange, repeat
from einops.layers.torch import Rearrange
from torch import nn


def pair(t):
    return t if isinstance(t, tuple) else (t, t)


class PreNorm(nn.Module):
    def __init__(self, in_channels: int, fn: nn.Module):
        super(PreNorm, self).__init__()
        self.norm = nn.LayerNorm(in_channels)
        self.fn = fn

    def forward(self, x, **kwargs):
        return self.fn(self.norm(x), **kwargs)


class FeedForward(nn.Module):
    def __init__(self,
                 in_channels: int,
                 hid_channels: int,
                 dropout: float = 0.):
        super(FeedForward, self).__init__()
        self.net = nn.Sequential(nn.Linear(in_channels, hid_channels),
                                 nn.GELU(), nn.Dropout(dropout),
                                 nn.Linear(hid_channels, in_channels),
                                 nn.Dropout(dropout))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.net(x)


class Attention(nn.Module):
    def __init__(self,
                 hid_channels: int,
                 heads: int = 8,
                 head_channels: int = 64,
                 dropout: float = 0.):
        super(Attention, self).__init__()
        inner_channels = head_channels * heads
        project_out = not (heads == 1 and head_channels == hid_channels)

        self.heads = heads
        self.scale = head_channels**-0.5

        self.attend = nn.Softmax(dim=-1)
        self.dropout = nn.Dropout(dropout)

        self.to_qkv = nn.Linear(hid_channels, inner_channels * 3, bias=False)

        self.to_out = nn.Sequential(
            nn.Linear(inner_channels, hid_channels),
            nn.Dropout(dropout)) if project_out else nn.Identity()

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        qkv = self.to_qkv(x).chunk(3, dim=-1)
        q, k, v = map(
            lambda t: rearrange(t, 'b n (h d) -> b h n d', h=self.heads), qkv)

        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale

        attn = self.attend(dots)
        attn = self.dropout(attn)

        out = torch.matmul(attn, v)
        out = rearrange(out, 'b h n d -> b n (h d)')
        return self.to_out(out)


class Transformer(nn.Module):
    def __init__(self,
                 hid_channels: int,
                 depth: int,
                 heads: int,
                 head_channels: int,
                 mlp_channels: int,
                 dropout: float = 0.):
        super(Transformer, self).__init__()
        self.layers = nn.ModuleList([])
        for _ in range(depth):
            self.layers.append(
                nn.ModuleList([
                    PreNorm(
                        hid_channels,
                        Attention(hid_channels,
                                  heads=heads,
                                  head_channels=head_channels,
                                  dropout=dropout)),
                    PreNorm(
                        hid_channels,
                        FeedForward(hid_channels, mlp_channels,
                                    dropout=dropout))
                ]))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        for attn, ff in self.layers:
            x = attn(x) + x
            x = ff(x) + x
        return x


class ArjunViT(nn.Module):
    r'''
    Arjun et al. employ a variation of the Transformer, the Vision Transformer to process EEG signals for emotion recognition. For more details, please refer to the following information. 

    It is worth noting that this model is not designed for EEG analysis, but shows good performance and can serve as a good research start.

    - Paper: Arjun A, Rajpoot A S, Panicker M R. Introducing attention mechanism for eeg signals: Emotion recognition with vision transformers[C]//2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE, 2021: 5723-5726.
    - URL: https://ieeexplore.ieee.org/abstract/document/9629837

    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',
                    offline_transform=transforms.Compose([
                        transforms.MeanStdNormalize(),
                        transforms.To2d()
                    ]),
                    online_transform=transforms.Compose([
                        transforms.ToTensor(),
                    ]),
                    label_transform=transforms.Compose([
                        transforms.Select('valence'),
                        transforms.Binary(5.0),
                    ]))
        model = ArjunViT(chunk_size=128,
                         t_patch_size=50,
                         num_electrodes=32,
                         num_classes=2)

    Args:
       num_electrodes (int): The number of electrodes. (default: :obj:`32`)
        chunk_size (int): Number of data points included in each EEG chunk. (default: :obj:`128`)
        t_patch_size (int): The size of each input patch at the temporal (chunk size) dimension. (default: :obj:`32`)
        patch_size (tuple): The size (resolution) of each input patch. (default: :obj:`(3, 3)`)
        hid_channels (int): The feature dimension of embeded patch. (default: :obj:`32`)
        depth (int): The number of attention layers for each transformer block. (default: :obj:`3`)
        heads (int): The number of attention heads for each attention layer. (default: :obj:`4`)
        head_channels (int): The dimension of each attention head for each attention layer. (default: :obj:`8`)
        mlp_channels (int): The number of hidden nodes in the fully connected layer of each transformer block. (default: :obj:`64`)
        num_classes (int): The number of classes to predict. (default: :obj:`2`)
        embed_dropout (float): Probability of an element to be zeroed in the dropout layers of the embedding layers. (default: :obj:`0.0`)
        dropout (float): Probability of an element to be zeroed in the dropout layers of the transformer layers. (default: :obj:`0.0`)
        pool_func (str): The pool function before the classifier, optionally including :obj:`cls` and :obj:`mean`, where :obj:`cls` represents selecting classification-related token and :obj:`mean` represents the average pooling. (default: :obj:`cls`)
    '''
    def __init__(self,
                 num_electrodes: int = 32,
                 chunk_size: int = 128,
                 t_patch_size: int = 32,
                 hid_channels: int = 32,
                 depth: int = 3,
                 heads: int = 4,
                 head_channels: int = 64,
                 mlp_channels: int = 64,
                 num_classes: int = 2,
                 embed_dropout: float = 0.,
                 dropout: float = 0.,
                 pool_func: str = 'cls'):
        super(ArjunViT, self).__init__()
        self.num_electrodes = num_electrodes
        self.chunk_size = chunk_size
        self.t_patch_size = t_patch_size
        self.hid_channels = hid_channels
        self.depth = depth
        self.heads = heads
        self.head_channels = head_channels
        self.mlp_channels = mlp_channels
        self.num_classes = num_classes
        self.embed_dropout = embed_dropout
        self.dropout = dropout
        self.pool_func = pool_func

        assert chunk_size % t_patch_size == 0, f'EEG chunk size {chunk_size} must be divisible by the temporal patch size {t_patch_size}.'

        num_patches = chunk_size // t_patch_size
        patch_channels = num_electrodes * t_patch_size

        assert pool_func in {
            'cls', 'mean'
        }, 'pool_func must be either cls (cls token) or mean (mean pooling)'

        self.to_patch_embedding = nn.Sequential(
            Rearrange('b c (w p) -> b w (c p)', p=t_patch_size),
            nn.Linear(patch_channels, hid_channels),
        )

        self.pos_embedding = nn.Parameter(
            torch.randn(1, num_patches + 1, hid_channels))
        self.cls_token = nn.Parameter(torch.randn(1, 1, hid_channels))
        self.dropout = nn.Dropout(embed_dropout)

        self.transformer = Transformer(hid_channels, depth, heads,
                                       head_channels, mlp_channels, dropout)

        self.pool_func = pool_func

        self.mlp_head = nn.Sequential(nn.LayerNorm(hid_channels),
                                      nn.Linear(hid_channels, num_classes))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        r'''
        Args:
            x (torch.Tensor): EEG signal representation, the ideal input shape is :obj:`[n, 32, 128]`. Here, :obj:`n` corresponds to the batch size, :obj:`32` corresponds to :obj:`num_electrodes`, and :obj:`chunk_size` corresponds to :obj:`chunk_size`.

        Returns:
            torch.Tensor[number of sample, number of classes]: the predicted probability that the samples belong to the classes.
        '''
        x = self.to_patch_embedding(x)
        x = rearrange(x, 'b ... d -> b (...) d')
        b, n, _ = x.shape

        cls_tokens = repeat(self.cls_token, '1 1 d -> b 1 d', b=b)
        x = torch.cat((cls_tokens, x), dim=1)
        x += self.pos_embedding[:, :(n + 1)]
        x = self.dropout(x)

        x = self.transformer(x)

        x = x.mean(dim=1) if self.pool_func == 'mean' else x[:, 0]

        return self.mlp_head(x)