The technology relates to bandwidth constrained communication systems with neural network based detection. In some embodiments, a bandwidth constrained equalized transport (BCET) communication system comprises: a transmitter comprising an error control code encoder, a pulse-shaping filter, and a first interleaver; a communication channel; and a receiver comprising a neural network processing block that processes a received signal. The error control code encoder can append redundant information onto the signal. The pulse-shaping filter can intentionally introduce memory into the signal in the form of inter-symbol interference. The first interleaver can change a temporal order of the symbols in the signal. The neural network can be trained with positive mappings between transmitted and decoded training signals, or negative mappings between training signals and erroneous decoded signals that are known to contain errors.

Methods and apparatus for use in communication systems using recursive modulation schemes with a Low Density Generator Matrix code (including an irregular repeat accumulate (IRA) code) are described that have reduced complexity and thus reduced cost compared to prior art systems. A communication system is described in which the transmitter concatenates a low density generator matrix code with an accumulator followed by a recursive modulator in order to eliminate the use of an interleaver, and in which the receiver combines the decoder for the accumulator and the soft demodulator into a single joint decoder in order reduce the number of components and complexity. Another variation is also described in which the transmitter is further simplified by eliminated the accumulator altogether, and in which the receiver is further simplified by replacing the joint decoder with a soft demodulator prior to the LDGM soft decoder.