Computing Nonlinear Functions Over Gaussian Multiple-Access Channel with Feedback

In the literature, it has already been shown that channel feedback helps to construct low-complexity and high reliability (LCHR) computing schemes over the Gaussian multiple-access channel (GMAC). However, only linear functions are addressed in these existing schemes. In this paper, we show that for the GMAC, channel feedback is also useful for designing LCHR schemes for computing non-linear functions, such as polynomial and absolute value functions. The relationship between the computation rate and the receiver’s computation distortion is characterized, and numerical results show that for given distortion, our proposed schemes may achieve larger computation rates than those schemes without using feedback for some cases.