Task-Focused Dynamic Network Inference from Node Attribute Time Series

NetSciX 2026 Conference Submission80 Authors

Keywords: dynamic networks, network inference, dynamic attributed network simulation, graph neural network

TL;DR: DyNRI is a method that infers dynamic graphs from node-attribute time series using an encoder–decoder with Gumbel-Softmax, optimizing the inferred structure for multi-step prediction and evaluated via a co-evolution simulation.

Submission Number: 80