Deep Learning Explorations in Wireless Channel Encoders

The modern wireless mobile network architectures are diverse and complex and face huge capacity demands. With increase in demand, monitoring and management of the multitude of network elements has become intractable. To meet the end-user requirements, Internet Service Providers (ISPs) are developing intelligent heterogeneous architectures and approaches that can offspring the 5th generation of mobile systems (5G). Thus, developing mathematical models and adapting them to modems has become far more difficult with 5G. Channel models are important because wireless PHYs must encode, transmit, receive, and decode data in such a way that bit-error-rates (BERs) stay acceptably low. Embedding versatile machine intelligence and deep learning to develop such mathematical models into future mobile networks is gaining much interest of researchers worldwide. In this paper, it is shown that Deep learning models can take transmitted and received baseband data and uses the results to optimize wireless channel encoders in mobile networks. The data fed as input to deep learning can come from channel measurements. Due to availability of data in large amount, deep learning can learn and produce coefficients in more accurate manner. The paper also shows that heterogeneous information received from 5G can be extracted efficiently for getting progressive abstract correlations from the data, while decreasing the effort done in pre-processing.