DQfD-AIPT: An Intelligent Penetration Testing Framework Incorporating Expert Demonstration Data

Table 6

Special hyperparameter setting of the DQfD.

Hyperparameter	Value
Pretrain step	1000
N-step return weight	1.0
Supervised loss weight	1.0
L 2 regularisation weight	1.0
Expert margin	0.8
N of N-step return	10
Prioritized replay exponent	0.4
Prioritized replay constants	0.001
Prioritized replay constants	1.0
Prioritized replay importance sampling exponent	0.6