A Deep Learning-Based User Selection Scheme for Cooperative NOMA System with Imperfect CSI

Algorithm 2

Proposed learning network.

Input: Data .

Output: The estimated output signal vectors.

1: while There is a training sequence of length do

2: Transmits features to the fourth layer.

3: Use the dropout function on the network.

4: Two parallel LSTM layers learn uncorrelated CSI respectively.

5: Two dense layers correspond to the two decoding in SIC process.

6: while The loss function does not converge do

7:  Use the RMSprop optimizer.

8: if Network converges then

9: Feed data for the current slot into LSTM network.

10: if CSI is successfully identified then

11:  Select the best near user.

12: else

13:  Retrain data.

14: else

15: Return to step 1.

16: while Find the optimal selection do

17: Output the estimated output signal vectors.