DANI: A Fast Diffusion Aware Network Inference Algorithm

Abstract

The fast growth of social networks and their privacy requirements in recent years, has lead to increasing difficulty in obtaining complete topology of these networks. However, diffusion information over these networks is available and many algorithms have been proposed to infer the underlying networks by using this information. The previously proposed algorithms only focus on inferring more links and do not pay attention to the important characteristics of the underlying social networks In this paper, we propose a novel algorithm, called DANI, to infer the underlying network structure while preserving its properties by using the diffusion information. Moreover, the running time of the proposed method is considerably lower than the previous methods. We applied the proposed method to both real and synthetic networks. The experimental results showed that DANI has higher accuracy and lower run time compared to well-known network inference methods.

DANI: Fast Diffusion Aware Network Inference with Preserving Topological Structure Property

“…Due to its sampling method, its convergence is slow. PBI first runs DANI [26] to achieve a pairwise interaction network and converts it to an undirected network, then iteratively employs the TNT sampler and MCMC algorithm to add or remove a random edge to this network [7]. Because of using DANI's output as the initial graph, it preserves modular structure and can detect edges in undirected graphs with high accuracy after Bayesian iterations.…”

When Less Is More: Systematic Analysis of Cascade-Based Community Detection

“…KernelCascade (Du et al, 2012) also extends NetRate, here the authors use a less restrictive cascade model. The DANI algorithm (Ramezani et al, 2017) is interesting because it explicitly accounts for the community structure to enhance the inference of networks' edges. There are some other network inference algorithms not covered here, see, e.g., (Daneshmand et al, 2014;Gomez-Rodriguez et al, 2013a;Netrapalli and Sanghavi, 2012;Saito et al, 2009;Silva et al, 2011;Snowsill et al, 2011;Wang et al, 2012;Yang and Zha, 2013;Zhou et al, 2013).…”

“…As an input, similarity scores of node pairs are computed, based on their joint participation in cascades. The R-CoDi algorithm in (Ramezani et al, 2018) starts with a random partition, while D-CoDi starts with a partition obtained by DANI (Ramezani et al, 2017). We use all four mentioned algorithms as our baselines.…”

“…Another example is Twitter, which sells the stream of tweets to other companies, but the social graph is not disclosed; for each retweet, only information about the root node of the cascade is available. The problem of hidden network inference received much attention recently (Daneshmand et al, 2014;Gomez-Rodriguez and Schölkopf, 2012;Ramezani et al, 2017). While these papers aim to recover the actual network connections, in many cases such detailed information may not be needed.…”

When Less Is More: Systematic Analysis of Cascade-Based Community Detection

“…Another example is Twitter, which sells a stream of tweets to other companies, but the social graph is not disclosed; for each retweet, only information about the cascade's root node is available. The hidden network inference problem received much attention recently [12,20,24,55]. While these articles aim at recovering the actual network connections, such detailed information may not be needed in many cases.…”