A Deep Learning-Based Power Control and Consensus Performance of Spectrum Sharing in the CR Network

Algorithm 1

Power control policy-deep learning training based.

  Initialize D with capacity O

  Initialize network with variable weights

  Initialize and and then obtain (1)

  For , do

  Update via power control policy of PU (2) or (3)

 With iterations , choose an arbitrary action ; otherwise, select

  Obtain via arbitrary model (5) and detect reward

  Store transition in

  Sensing delay

  Repeat sensing delay

  If then

  Sample random minibatch of iterations from ,

  Here, the index is uniformly selected at independent

  Minimize loss function of 12, in which goal is given by (15)

  Adjust

  End if

   is the target state and then initialize and and then gain

 end if

  end for