Декодирование поведенческих намерений

import torch
import torch.nn as nn

class EEGNet(nn.Module):
    def __init__(self, num_classes=4):
        super(EEGNet, self).__init__()
        
        # Первый слой - временная свертка
        self.temporal_conv = nn.Conv1d(64, 16, kernel_size=64, stride=2)
        self.batch_norm1 = nn.BatchNorm1d(16)
        self.activation1 = nn.ELU()
        self.dropout1 = nn.Dropout(0.25)
        
        # Второй слой - пространственная свертка
        self.spatial_conv = nn.Conv1d(16, 32, kernel_size=16, groups=16)
        self.batch_norm2 = nn.BatchNorm1d(32)
        self.activation2 = nn.ELU()
        self.dropout2 = nn.Dropout(0.25)
        
        # Классификатор
        self.classifier = nn.Sequential(
            nn.Linear(32 * 62, num_classes),
            nn.Softmax(dim=1)
        )
    
    def forward(self, x):
        x = self.dropout1(self.activation1(self.batch_norm1(self.temporal_conv(x))))
        x = self.dropout2(self.activation2(self.batch_norm2(self.spatial_conv(x))))
        x = x.view(x.size(0), -1)
        return self.classifier(x)