Effect of clustering property on complex network reconstruction via compressed sensing

Complex networks are widely used to describe the interactions of real systems such as technological, social and biological systems. Compressed sensing method is one of the most effective data-driven methods which has been used to reconstruct the underlying structure of network from small amounts of measurement data. Although the compressed sensing-based methods show a powerful reconstructing ability for many kinds of networks, such as small-world network, scale-free network and so on, few works have taken the statistical properties of complex network into account. In fact, the statistical properties, such as clustering coefficient, have significant effect on network structure as well as the measurement data collected from the complex networked system. Thus, we investigate the relationship between clustering property and accuracy of network reconstruction based on small-world networks in this paper. Extensive experiments and analyses show that a more loosely distributed structure with low clustering property is more conducive than a compact structure with high clustering property to network reconstruction via compressed sensing.