Uncovering Patterns of Participant-Invariant Influence in Networks

Abstract

In this paper, we explore the nature of influence in a network. The concept of participant-invariant influence is derived from an influence matrix M specifically designed to explore this phenomenon. Through nonnegative matrix factorization approximation, we managed to extract a participant-invariant matrix H representing a shared pattern that all participants must obey. The acquired H is highly field-related and can be further utilized to cluster factual networks. Our discovery of the unveiled participant-independent influence within network dynamics opens up new avenues for further research on network behavior and its implications.