Table 5.

Hyperparameters for the GraphATT-DTA model.

Hyperparameter	Setting
Graph neural network layers	2 or 3 or 5
K size	8
Epoch	1000
Batch size	32
Learning rate	0.0005
Optimizer	Adam
Weight decay	0.00001
Embedding size of a feature	128
Fully connected layers	(1024, 512, 1)
Dropout	0.1 or 0.2
Early stopping	30