All-Cause Death Prediction Method for CHD Based on Graph Convolutional Networks

Table 3

Parameter configuration of model.

Parameter name	Parameter value	Parameter description
Epochs	300	Training batch size
lr	0.01	Learning rate
weight_decay	5e-4	Weight decay
K	4	Number of neighbors
nhid1	16	Number of hidden layers 1
nhid2	8	Number of hidden layers 2
Dropout	0.5	Drop rate
Beta	5e-10	Loss function parameter 1
Theta	0.001	Loss function parameter 2
Seed	21	Random number seed
Patience	40	Early stop rounds