Budget-Constrained Robust Influence Maximization
Abstract
Departing from traditional combinatorial, independent cascade influence maximization, we propose a continuous, correlation-robust influence maximization model. Instead of a deterministic seeding of nodes, a budgeted selection of discounts is now used to affect the likelihood of seeding. Additionally, edge probabilities are no longer assumed to be independent and are instead coupled adversarially. This model features a combination of increased computational tractability while also providing some means to express more sophisticated edge relationships or dependencies. We provide a study of the maximization problems, and show favorable performance of its solutions as compared to those of previous works assuming independence. More precisely, we measure the relative trade-off in performance between independent cascade and adversarial models. Further, we show that this proposed model can be used for networks with variable node rewards. We conclude with experiments on real-world datasets.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2023
- Accession Number
- AD1212885
Entities
People
- Jeremy L. Berg
Organizations
- Naval Postgraduate School