Nonlinear equalization was previously shown to be beneficial for recording channels with nonlinearities and signal-dependent noise. In this work, we study two types of simplified nonlinear filter architectures that are suitable for VLSI implementation. The performance loss due to the simplifications are not significant in most cases, compared to the signal-to-noise ratio gain achieved by nonlinear equalization. In one form of simplification the need for full multiplications is eliminated and in the other the number of multiplications in the original nonlinear filter architecture is reduced significantly. Work on experimental and simulated data are presented.