Deep Learning-Based Solutions for 5G Network and 5G-Enabled Internet of Vehicles: Advances, Meta-Data Analysis, and Future Direction

Table 4

The summary of deep learning algorithm characteristics and learning suitability in 5G.

Deep learning algorithm	Learning suitability	Application in 5G
DRL	Reinforcement	Energy consumption minimization, beam selection, load balancing scheme, distributed offloading framework, service function chaining mapping, packet drop probability and resource utilization, network slicing scheme; vehicles platooning management, optimal locations deployment, time division duplex resource allocation; initial window decision policy, quality of service, maximizing energy efficiency, signal to interference plus noise ratio, improve data rate at the edge and reduces power transmitted
CNN	Supervised	Millimeter wave positioning, channel estimation, scalable caching scheme, signal interference detector, denial-of-service detector scheme, predict user location, detecting 5G signal
DDNN	Supervised learning	Pilot assignment in massive MIMO, power factor allocator, slices resources predictor, 3D base station positioning system, beam selector, authentication scheme
Hybrid algorithm	Supervised/unsupervised	Mobile network traffic loads forecasting, channel state information prediction, power transmission, cyber security system
Autoencoder	Unsupervised/semisupervised	Deep-SCMA scheme, cache scheme
LSTM	Supervised	Holistic handover management, intelligent cache scheme; TV broadband allocation, hotspot virtual small cell allocator, discontinue reception prediction, predict traffic flow
GAN	Semisupervised and unsupervised learning	5G-enabled drone monitoring system; slice resource management