Uncovering Multiple Diffusion Networks Using the First-Hand Sharing Pattern

Liao, Pei-Lun; Chou, Chung‐Kuang; Chen, Ming-Syan

doi:10.1137/1.9781611974348.8

Cited by 6 publications

(11 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wang et al (MMRate [27]) and Yang et al (MixCascades [30]) were among the first to build a general framework that infers multilayer diffusion networks solely from spreading data. Later, Liao et al [16] proposed FASTEN that improves the inference accuracy by incorporating a decay parameter in the diffusion model. Another relevant line of research uses the Marked Multivariate Hawkes Process [17] to infer multilayer diffusion networks.…”

Section: Related Workmentioning

confidence: 99%

“…However, the range of synthetic test settings, as reported in the papers, is often limited. For example, the tests are usually performed on synthetic networks with fixed numbers of nodes, edges, and layers [16,27,30]. These settings also likely differ from real-world situations in, for example, having independently generated network layers and the assumption that any item spreads on either one layer or another.…”

Section: Related Workmentioning

confidence: 99%

“…In the first set of experiments, we have the setting of 𝑁 = 1000, 𝜇 𝑖𝑛 = 0.5, 𝐾 = 2, 𝜙 = 0, |𝐸 𝐴 | = 4422, 𝜖 𝑚𝑎𝑥 = 0, no cascade filtering, and 𝛾 taking 1, 2, 4, or 8. Under each 𝛾 value, we inspect how the inference performance changes as we increase the number of cascades, more specifically when the cascade-edge ratio (i.e., the ratio between the number of cascades 𝐶 and the number of edges in the aggregated network |𝐸 𝐴 |, abbreviated as C-E ratio) is 1, 2, 4, 8, or 16.…”

Section: Varying Cascade Size Distributionmentioning

confidence: 99%

“…Varying level of mixed spreading. In the final set of experiments, we have the setting of 𝑁 = 1000, 𝐾 = 2, 𝜙 = 0, 𝜇 𝑖𝑛 = 0.5, 𝛾 = 2, 𝑠 𝑐 = 8, 𝜖 𝑚𝑎𝑥 taking 0, 0.2, or 0.4, and C-E ratio taking 1, 2, 4, 8, or 16. We find that the accuracy of both edge existence and edge weight inference does not differ signficantly with the level of mixed spreading, but the accuracy of the cascade layer membership inference indeed decreases with layer mixing, as shown in Figure 3g.…”

Section: 27mentioning

confidence: 99%

“…This heterogeneity can be best represented with a multilayer network [15] where node existence, edge connectivity, and edge weights can differ across layers. Under this framework, researchers have developed methods for inferring multilayer diffusion networks from spreading data [11,12,16,25,27,30].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Applicability of Multilayer Diffusion Network Inference to Social Media Data

Xia¹,

Chen²,

Kivelä³

2021

Preprint

View full text Add to dashboard Cite

Information on social media spreads through an underlying diffusion network that connects people of common interests and opinions. This diffusion network often comprises multiple layers, each capturing the spreading dynamics of a certain type of information characterized by, for example, topic, attitude, or language. Researchers have previously proposed methods to infer these underlying multilayer diffusion networks from observed spreading patterns, but little is known about how well these methods perform across the range of realistic spreading data. In this paper, we first introduce an effective implementation of the inference method that can achieve higher accuracy than existing implementations in comparable runtime. Then, we conduct an extensive series of synthetic data experiments to systematically analyze the performance of the method, under varied network structure (e.g. density, number of layers) and information diffusion settings (e.g. cascade size, layer mixing) that are designed to mimic real-world spreading on social media. Our findings include that the inference accuracy varies extremely with network density, and that the method fails to decompose the diffusion network correctly when most cascades in the data reach a limited audience. In demonstrating the conditions under which the inference accuracy is extremely low, our paper highlights the need to carefully evaluate the applicability of the method before running the inference on real data. Practically, our results serve as a reference for this evaluation, and our publicly available implementation supports further testing under personalized settings.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Varying Cascade Size Distributionmentioning

confidence: 99%

Section: 27mentioning

confidence: 99%