CSL-SFNet for Cooperative Spectrum Sensing in Cognitive Satellite Network with GEO and LEO Satellites

Algorithm 1

CSL-SFNet Based CSS.

1: Set ; initialize maximum training epochs ; initialize and with random weights;

2: Collect the training set and feed it into CSLNet;

3: fordo

4:  According to Equation (15), SUs train the CSLNet by the backpropagation algorithm to obtain the optimal model parameters ;

5: end for

6: SUs send the output SSFs of well-trained CSLNet to FC;

7: FC combines SSFs as ;

8: fordo

9:  FC trains the Soft-FusionNet by the backpropagation algorithm to get the optimal model parameters ;

10: end for

11: FC calculates the threshold based on Equations (18) and (19);

12: SUs receive the test sample online and input it into the well-trained CSLNet to obtain the new SSFs, which are transmitted to FC;

13: FC outputs the corresponding category probability vector and and decides the final PU state based on Equation (20);

14: FC broadcasts the final decision result to each SU.