A new neural equalizer for decision-feedback equalization

In this paper, we propose a simple but powerful neural decision-feedback equalizer trained with a fast-converging adaptive filter algorithm, which is efficient, simple, and numerically robust. The equalizer can be viewed as a neuron with fixed or adaptive sigmoidal nonlinearity. The simulation results on various time-invariant and time-varying channel equalization benchmarks have shown its surprisingly good performance compared to other sophisticated neural network-based equalizers. The empirical results have demonstrated the potential values of the proposed neuronal equalizers in practice