2019/04/17 by Vishnu Raj, Sheetal Kalyani, Raj, Vishnu +1
Computer Science · #FOS: Computer and information sciences #Information Theory (cs.IT) #Machine Learning (cs.LG) #Wireless Signal Modulation Classification
paper · pdf · doi:10.48550/arxiv.1904.08559
openalex publication_date 2019/04/17 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
Recent research in the design of end to end communication system using deep\nlearning has produced models which can outperform traditional communication\nschemes. Most of these architectures leveraged autoencoders to design the\nencoder at the transmitter and decoder at the receiver and train them jointly\nby modeling transmit symbols as latent codes from the encoder. However, in\ncommunication systems, the receiver has to work with noise corrupted versions\nof transmit symbols. Traditional autoencoders are not designed to work with\nlatent codes corrupted with noise. In this work, we provide a framework to\ndesign end to end communication systems which accounts for the existence of\nnoise corrupted transmit symbols. The proposed method uses deep neural\narchitecture. An objective function for optimizing these models is derived\nbased on the concepts of variational inference. Further, domain knowledge such\nas channel type can be systematically integrated into the objective. Through\nnumerical simulation, the proposed method is shown to consistently produce\nmodels with better packing density and achieving it faster in multiple popular\nchannel models as compared to the previous works leveraging deep learning\nmodels.\n