Towards a Cognitive Millimeter-wave Radio-over-Fiber Centralized Radio Access Network for Future Generation Small Cell Deployment

Fiber-wireless access technology is an active research topic for future generation communication networks. However, the nonlinear distortion of fiber can cause problems for higher-order modulation signals. The use of shorter wavelength radio frequency signals in 5G leads to requiring more base stations, which would increase network management complexity. This thesis explores the use of deep learning in coordinated multipoint transmission, and develops novel phase noise tolerant receivers that use incoherent detection and deep learning algorithms. Research findings demonstrate the ability of deep learning algorithms and the possibility to implement deep learning algorithms in future generation cognitive-communication networks.