On the Construction of Neural Networks via Wireless Ad Hoc Networks

Abstract

Due to the similarities between neural networks and ad hoc networks, ad hoc networks can be applied to construct distributed neural network to perform complex computation. In this paper, the broadcasting nature of wireless signal and communication interactions among the nodes in ad hoc network are applied to construct neural network, which reveals the interplay between communication and computing. In the constructed neural network, the neuron is the wireless node. The communication interactions are applied to increase the number of hidden layers. The constructed neural network is further applied in target positioning and its accuracy is verified by simulation results. This paper shows that the wireless networks can be applied in computation, which may motivate the construction of large-scale neural networks via the wireless networks.