Large Scale Network Structure Inference Algorithm Based on Probabilistic Graph Modeling

Abstract

Interactions between individuals or entities exhibit dynamic characteristics, often depicted through a series of network snapshots, capturing interactions within short timeframes. Central to analyzing these evolving networks is the task of change-point detection. This involves pinpointing moments when the overarching interaction pattern undergoes significant shifts and measuring the scale and nature of these changes.. Thus, they develop a strategy that consistently solves the network change-point detection issue and formalise it inside an online probabilistic learning framework. The proposed method is Equivariant Quantum Neural Networks (EQNN). This method combines a user-defined parameter that specifies a goal false-positive rate with a flexible hierarchical random graph model that uses the Bayesian hypothesis test. EQNN method is used to change-point detection and aiming to identify changes in the large scale structure of evolving networks. The proposed method shows better existing method like graph neural network (GNN), Convolution Neural Network (CNN) and Multilayer Perception Neural Network (MPNN). The proposed method has error value of 0.5% which is lower than that of GNN, CNN and MPNN whose error values are 0.9%, 1.3% and 1.7% respectively.