Localization of diffusion sources in complex networks: A maximum-largest method

Networks play a role that interactions through which behaviors and diseases can spread. Identifying or locating all the sources in a large network is an important step towards understanding the transmission mechanism. Based on the network structure and backward diffusion-based method, we propose a maximum-largest method to locate sources with limited observers. Results of applying this method to modeling networks and empirical networks demonstrate that our method is superior on a larger networks size for a certain fraction of observers. Besides, our method is very robust for different strategies of choosing observers. Furthermore, the performance of our method is better than the previous method (the maximum–minimum method), especially for a small fraction of available observers. What is more, the performance of our method can be further improved by virtue of Gaussian kernel, which is very robust against noise case. Our analysis provides a route for improving source localization in large networks. •An efficient approach to locate multiple sources is proposed.•Our method has a better performance on a larger network given a certain fraction of observers.•The performance of our method can be further improved by virtue of Gaussian kernel.•Compared with the MMM, our method is more robust to placements of observers and noise.