Sequential decision based learning method for influence maximization

DOI10.1016/J.TCS.2025.115147MaRDI QIDQ7006269zbMATH OpenFDO

Authors Zizhen Zhang, Deying Li, Yongcai Wang, Wenping Chen, Yuqing Zhu

Publication date 31 March 2025

Published in Theoretical Computer Science (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/J.TCS.2025.115147

zbMATH Keywords

generative flow networkshypergraphinfluence maximizationsubmodular reinforcement learning

Mathematics Subject Classification ID

Markov and semi-Markov decision processes (90C40) Decision theory (91B06) Social networks; opinion dynamics (91D30)

Review text

The paper studies sequential decision based learning method for influence maximization. It presents a sequential decision-based learning approach for tackling the influence maximization problem through a data-driven end-to-end methodology. By sampling from a target distribution where energy aligns with a pre-defined score function, generative flow network for influence maximization captures seed sets with consistent outcomes across different node sequences improving solution effectiveness. The paper proposes a simple, policy-gradient-based algorithm that handles sub-modular rewards by maximizing marginal gains. Numerical simulations are conducted to show the performance of the model.

Cites work

• Adaptive submodularity: theory and applications in active learning and stochastic optimization
• An analysis of approximations for maximizing submodular set functions—I
• Automata, Languages and Programming
• Maximizing social influence in nearly optimal time
• The book of why. The new science of cause and effect

This page was built for publication: Sequential decision based learning method for influence maximization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7006269)