Influence Maximization in Multiagent Systems by a Graph Embedding Method: Dealing With Probabilistically Unstable...

Area of Science:

• Network Science
• Multiagent Systems
• Graph Theory

Background:

• The influence maximization problem is crucial for targeted information dissemination.
• Networks with probabilistically unstable links present unique challenges for traditional influence maximization approaches.
• Multiagent systems offer a framework for modeling complex network dynamics.

Purpose of the Study:

• To develop novel diffusion models for influence maximization in networks with probabilistically unstable links.
• To establish a multiagent system model for influence maximization considering link instability.
• To propose a graph embedding method for identifying optimal seed sets in such networks.

Main Methods:

• Design of two diffusion models: unstable-link independent cascade (UIC) and unstable-link linear threshold (ULT).
• Establishment of a multiagent system (MAS) model with interaction rules for probabilistically unstable links (PULs).
• Development of the unstable-similarity2vec (US2vec) graph embedding approach to capture node structural similarity.

Main Results:

• The proposed UIC and ULT models effectively capture diffusion dynamics in networks with PULs.
• The US2vec method accurately embeds network structures, reflecting node instability.
• The developed algorithm successfully identifies seed sets for influence maximization using US2vec embeddings.

Conclusions:

• The study validates the proposed models and algorithms for influence maximization in networks with PULs.
• The US2vec approach provides an effective graph embedding solution for identifying optimal seed sets.
• The findings offer insights into optimizing influence spread in dynamic and uncertain network environments.