Unsupervised Attributed Dynamic Network Embedding with Stability Guarantees

2025-03-04

AI Summary

For dynamic networks with time-varying node attributes, this paper proposes AUASE, an unsupervised embedding framework designed to ensure cross-temporal stability of node representations—i.e., structurally and behaviorally consistent nodes receive identical embeddings across time steps, enabling meaningful temporal comparability. AUASE is the first unsupervised method to theoretically guarantee embedding stability, proving uniform convergence to the latent position model—a theoretical gap previously unaddressed in attributed dynamic graph representation learning. Methodologically, AUASE integrates unfolded adjacency spectral embedding, temporal attribute fusion, and statistical consistency analysis to jointly model the coupling between structural evolution and time-varying attribute covariates. Extensive experiments on three real-world datasets demonstrate that AUASE significantly outperforms state-of-the-art dynamic graph embedding methods on both link prediction and node classification tasks.

Technology Category

• Planning, Routing, and Scheduling: Temporal Planning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Graph-based Machine Learning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Computational social science

Abstract

Stability for dynamic network embeddings ensures that nodes behaving the same at different times receive the same embedding, allowing comparison of nodes in the network across time. We present attributed unfolded adjacency spectral embedding (AUASE), a stable unsupervised representation learning framework for dynamic networks in which nodes are attributed with time-varying covariate information. To establish stability, we prove uniform convergence to an associated latent position model. We quantify the benefits of our dynamic embedding by comparing with state-of-the-art network representation learning methods on three real attributed networks. To the best of our knowledge, AUASE is the only attributed dynamic embedding that satisfies stability guarantees without the need for ground truth labels, which we demonstrate provides significant improvements for link prediction and node classification.

Problem

• Ensures stable embeddings for nodes in dynamic networks over time.
• Develops AUASE for unsupervised learning in attributed dynamic networks.
• Improves link prediction and node classification without ground truth labels.

Innovation

• AUASE: stable unsupervised dynamic network embedding
• Uniform convergence to latent position model
• No ground truth labels for stability guarantees