Joint Optimization of Jamming Link and Power Control in Communication Countermeasures: A Multiagent Deep Reinforcement Learning Approach

Table 3

Main hyper parameters of MASAC.

Parameters	Value
	-0.25
	5
	1
Training episodes ()	5000
Total steps of each episode ()	500
Soft updating rate ()	0.01
Capacity of CRB	2^17
Minibatch size ()	256
Initial value of entropy coefficient ()	1
Discount factor ()	0.98
Learning rate of policy network	0.001
Learning rate of evaluation network	0.003
Training threshold of CRB ()	1024