Coded Wireless Distributed Computing via Interference Alignment

This paper proposes a coded parallel computing scheme (CPC) for the wireless MapReduce system where multiple nodes simultaneously exchange information via a wireless interference network. The CPC scheme is based on the coded distributed computing scheme proposed by Li et al., followed by interference alignment to cancel interference caused by the concurrent transmission. It is shown that CPC can significantly reduce the communication latency by jointly exploiting the parallel transmission and coded multicasting opportunities. Different from the previous coded computing schemes whose communication latency increase with the total number of computing nodes K, the communication latency of CPC decreases with K when K is relatively large. As K tends to infinity, all previous schemes achieve constant communication latency, while CPC approaches zero communication latency.