Table 1 Notations and the corresponding interpretation.
Symbol | Interpretation |
|---|---|
Xi | Spatial-temporal feature data of the i-th sample |
Vt | Node feature matrix within window t |
Et | edge index matrix |
Wt | edge weight vector |
B | Batch size |
HG | Convolutional encoder output |
HT | Sequence encoder output |
E | Number of training epochs |
Lr | Learning rate of optimizer |
LG | GCN layers |
LT | Transformer encoder layers |
w | Sliding window size |
s | Sliding window step size |
k | Nearest neighbor count for node connections |
h | Number of multi-head attention heads |
dFNN | Feed-forward neural network hidden dimension |
yt | Window image corresponding to a label |
p | Dropout rate |
dG | Graph convolutional encoder output dimension |
dT | Temporal encoder output dimension |
dFC | Fully connected layer output dimension |
\(\:\tau\:\) | Similarity threshold for edge construction |
