Open Access
Table 4
Detailed structure of DG-MGNet.
| Module | Information on the structure |
|---|---|
| Temporal encoder | Conv1D(1→64, k=9, s=2) - Conv1D(64→128, k=9, s=2) - Conv1D(128→128, k=9, s=2) + {GroupNorm(8) + ReLU}; AdaptiveAvgPool1D(1); Linear(128→128) |
| Graph generator | Prior graph as an identity matrix; global residual graph as a learnable matrix; task graph generated via linear projections Wq/Wk: Linear(128→128, bias=False) ×2 |
| GraphConv | Two GCN layers: Linear(128→128, bias=False) |
| AttnPool | A learnable query vectorq (dim = 128); softmax-weighted summation over node outputs |
| Regression head | Linear (128,64), ReLU, Dropout (0.3), Linear (64,1) |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.