Skip to main content

Table 1 Configurations of the multi-view convolutional representation module in HybridAtt

From: A hybrid self-attention deep learning framework for multivariate sleep stage classification

TypeKernel sizeStridePadding
\(Conv1d_{1\_1}\)8×823
\(Conv1d_{1\_2}\)16×827
\(Conv1d_{1\_3}\)32×823
\(Conv1d_{1\_4}\)64×827
MaxPool1d1641
\(Conv1d_{2\_1}\)3×1611
\(Conv1d_{2\_2}\)5×1612
MaxPool1d2321
\(Conv1d_{3\_1}\)3×1611
\(Conv1d_{3\_2}\)5×1612
MaxPool1d3321
\(Conv2d_{1\_1}\)1×8×81,20,3
\(Conv2d_{1\_2}\)1×16×81,20,7
\(Conv2d_{1\_3}\)1×32×81,20,3
\(Conv2d_{1\_4}\)1×64×81,20,7
AvgPool2d11×61,40,1
\(Conv2d_{2\_1}\)3×3×161,11,1
\(Conv2d_{2\_2}\)5×5×161,12,2
AvgPool2d21×31,20,1
\(Conv2d_{3\_1}\)3×3×161,11,1
\(Conv2d_{3\_2}\)5×5×161,12,2
AvgPool2d314×314,20,1