Table 6.

Summary of the applications of ML for NOMA communications.

References	ML Model Architecture	Contributions	Remarks
[115]	Combined unsupervised and supervised learning	Spectrum sensing	Provided an accurate and effective spectrum sensing while maintaining optimal power allocation
[116]	LSTM-based DL models	Signal detection	Outperformed the (SIC) receiver and the limited radio resources
[117]	EE-CSL algorithm	Power optimization	Significantly minimized energy consumption for low computational complexity and achieved more significant sum rate than conventional MIMO orthogonal multiple access
[118]	DDQL-based RL algorithm	Transmission power optimization	Converged successfully in 91% of the test cases with a value better than the target, and performed better than SLSQP and TCONS algorithms
[119]	DREAM-FL scheme	Client selection	Provided more qualified clients with high model accuracy than FDMA- and TDMA-based solutions
[120]	LSTM-based DL algorithm	Channel coefficients prediction	Provided reliable performance even when cell capacity is increased