Intelligent Dynamic Spectrum Allocation in MEC-Enabled Cognitive Networks: A Multiagent Reinforcement Learning Approach

Table 2

Simulation parameters.

Parameters	Values
Number of PUs	1
Number of SUs	9
Number of MEC servers	1
Number of authorized channels	4
Discount factor	0.99
Learning rate
Replay buffer capacity	100 episodes of experience data
Batch size	16 episodes of experience data
Updating step	40 steps
Training epoch	2000 epochs
Training episode	100 episodes
Training slot	20 slots
Exploration probability	0.4 to 0.02