Neural Network Detection of Data Sequences in Communication Systems

We consider detection based on deep learning, and show it is possible to\ntrain detectors that perform well without any knowledge of the underlying\nchannel models. Moreover, when the channel model is known, we demonstrate that\nit is possible to train detectors that do not require channel state information\n(CSI). In particular, a technique we call a sliding bidirectional recurrent\nneural network (SBRNN) is proposed for detection where, after training, the\ndetector estimates the data in real-time as the signal stream arrives at the\nreceiver. We evaluate this algorithm, as well as other neural network (NN)\narchitectures, using the Poisson channel model, which is applicable to both\noptical and molecular communication systems. In addition, we also evaluate the\nperformance of this detection method applied to data sent over a molecular\ncommunication platform, where the channel model is difficult to model\nanalytically. We show that SBRNN is computationally efficient, and can perform\ndetection under various channel conditions without knowing the underlying\nchannel model. We also demonstrate that the bit error rate (BER) performance of\nthe proposed SBRNN detector is better than that of a Viterbi detector with\nimperfect CSI as well as that of other NN detectors that have been previously\nproposed. Finally, we show that the SBRNN can perform well in rapidly changing\nchannels, where the coherence time is on the order of a single symbol duration.\n