Reinforcement Learning for Routing in Cognitive Radio Ad Hoc Networks

Table 5

Simulation parameters and values for investigating the exploration approaches.

Category	Parameter	Value
SU	Traditional 𝜀-greedy exploration probability, 𝜀	{0.07, 0.14}
	Traditional softmax exploration temperature, 𝑀	{0.04, 0.05}
	Initial dynamic softmax temperature, 𝑀	0.05
	Dynamic softmax adjustment factor, 𝑓	0.01
	Dynamic softmax temperature range, [𝑀min,𝑀max]	[0.01, 0.1]
	Dynamic softmax 𝑄-value threshold, 𝜗	0.1
PU	Standard deviation of PUL, 𝜎PUL	{0.2, 0.8}
Channel	Mean PER, 𝜇PER	0
	Standard deviation of PER, 𝜎PER	0