Graph Similarity Based on Optimal Transmission for Optimal Deployment of Intelligent Wireless Sensor Networks

A graph similarity-based optimal deployment algorithm is proposed in this paper to find sensor nodes that can effectively recognize tasks in intelligent wireless sensor networks. A variational graph auto-encoder is used to encode each node as a multivariate normal distribution. Based on the optimal transport, we can transmute the similarity of two graphs into the Wasserstein distance between their learned multivariate normal distribution. We can obtain the importance of nodes for wireless sensor networks by measuring the Wasserstein distance of the standard and perturbed graphs. According to the importance ranking, the optimal sensor nodes can be selected to achieve task recognition. The results tested on CIMIS show that the sensor nodes selected by the proposed algorithm can achieve optimal deployment with high precision.