A Multiphase Semistatic Training Method for Swarm Confrontation Using Multiagent Deep Reinforcement Learning

Table 5

The main parameters of training.

Hyperparameters
Batch size	2048	Buffer size	20480	Learning rate	3.0e − 05	Beta	0.01
Epsilon	0.2	Lambda	0.95	Num epoch	3	Time horizon	128
Network setting				Reward signals
Hidden units	512	Num layers	3	Gamma	0.99	Strength	1