Table 3.

Selected hyper-parameters for the DPred model

Hyper-parameters	Selected values
Attention type	additive
Attention hidden size	32
Attention width	2
Number of filters	100
CNN stride size	2
Kernel size	2
Padding	same
Max-pooling stride-size	1
Max-pooling pool-size	2
Dropout ratio	0.1
Regularization rate	0.01
Dense size	100
Loss function	categorical_crossentropy
Optimizer	Adam
Learning rate	0.0001