Table 3 The details of Hyperparameters

GCN layer

8

Batch size

1

GCN hidden units

256

\(\alpha\) of GCN

0.7

Bi-LSTM layer

2

\(\beta _l\) of GCN

\(\log \left( \frac{\lambda }{l} + 1\right)\)

Bi-LSTM hidden units

32

\(\lambda\) of GCN

1.5

Learning rate

e-5

Dropout

0.2

Weight decay

0

The cutoff of distance map

14