Decentralized and Dynamic Band Selection in Uplink Enhanced Licensed-Assisted Access: Deep Reinforcement Learning Approach

Algorithm 2.

DQN training algorithm for dynamic band selection.

for each agent do

  Initialize replay buffer

  Initialize action value function with parameter

  Initialize target action value function with parameter

  Generate initial state from the environment simulator

end for

for do

  for each agent do

   Execute action from using -greedy policy

   Collect reward and observation

   Observe the next state from the environment simulator

   Store the transition into

   Sample random minibatch of transitions from

   Evaluate the target

   Perform a gradient descent step on with respect to

   Every C steps, update the target network according to

  end for

end for