Title: Rate Analysis and Deep Neural Network Detectors for SEFDM FTN Systems

Abstract: In this work we compare the capacity and achievable rate of uncoded faster than Nyquist (FTN) signalling in the frequency domain, also referred to as spectrally efficient FDM (SEFDM). We propose a deep residual convolutional neural network detector for SEFDM signals in additive white Gaussian noise channels, that allows to approach the Mazo limit in systems with up to 60 subcarriers. Notably, the deep detectors achieve a loss less than 0.4-0.7 dB for uncoded QPSK SEFDM systems of 12 to 60 subcarriers at a 15% spectral compression.