Multi-relational Network Autoregression Model with Latent Group Structures

2025-12-31 · Source: JMLR · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

Yimeng Ren, Xuening Zhu, Ganggang Xu, and Yanyuan Ma introduce the group tensor network autoregression (GTNAR) model, detailed in JMLR 27(78):1−135, 2026. This model addresses the challenge of quantifying effects in multi-relational networks and managing high-dimensional tensor-valued time series data. GTNAR models multiple network effects within an autoregressive framework by assuming distinct latent group structures for entities in each network. Within a given network, entities belonging to the same group share identical model parameters, which then vary across different networks. The researchers developed an iterative algorithm to simultaneously estimate both the model parameters and these latent group memberships. They theoretically demonstrate that group-wise parameters and memberships can be consistently estimated, even if the group numbers are over-specified. An information criterion is also provided for consistently selecting the correct group number for each network. The GTNAR method's utility is illustrated through its application to a Yelp dataset.

Key takeaway

For data scientists analyzing complex multi-relational time series, the GTNAR model offers a robust approach to uncover hidden group structures and their influence. You can apply this iterative algorithm to simultaneously estimate network parameters and latent group memberships, even when initial group number estimates are high. This method provides a principled way to quantify network effects in high-dimensional data, improving model interpretability and predictive accuracy. Consider using the provided information criterion to optimize group number selection for your specific datasets.

Key insights

GTNAR models multi-relational network effects in high-dimensional time series by inferring latent group structures and shared parameters.

Principles

• Entities in a group share parameters.
• Parameters differ across networks.
• Consistent estimation is achievable.

Method

An iterative algorithm simultaneously estimates group-wise model parameters and latent group memberships for each network, even with over-specified group numbers.

In practice

• Apply GTNAR to multi-relational time series.
• Use the information criterion for group number selection.
• Analyze Yelp-like datasets for network effects.

Topics

• Multi-relational Networks
• Network Autoregression
• Latent Group Structures
• Tensor Time Series
• Model Parameter Estimation
• Yelp Dataset

Code references

• MelodyRen1998/GTNAR
• JmlrOrg/v27

Best for: Research Scientist, AI Scientist, Data Scientist

Related on AIssential

• Nonparametric estimation of time-varying network connections by multi-stage smoothing — A multi-stage nonparametric smoothing estimator effectively models time-varying network edge probabilities with theoretical guarantees.
• TMR-GGNN: Credit Card Fraud Detection based on Time-Aware Multi-Relational Guided Graph Neural Network — TMR-GGNN leverages time-aware multi-relational GNNs and contrastive learning to improve credit card fraud detection, especially for rare cases.
• Structure and Relationships: Graph Neural Networks and a Pytorch Implementation — GNNs model interconnected data by transforming, aggregating, and updating node features based on neighbor relationships.
• From Uniform to Learned Graph Priors: Diffusion for Structure Discovery — Diffusion-parameterized adaptive prior calibrates latent graph distributions for reliable structure discovery.
• Graph Reduction in Multirelational Networks: A Spreading-Oriented Reduction Benchmark — Graph reduction's impact on influence maximization performance varies significantly by network type and reduction strategy.
• Graph Neural Networks Applications Across Domains: All Insights You Need — GNNs are default for relational data, but their real-world utility and deployment success vary significantly across domains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by JMLR.