Global graph features unveiled by unsupervised geometric deep learning

2024-10-04

Emerging Topics in Artificial Intelligence (ETAI) 2024

AI Summary

Modeling and classifying highly variable complex systems—such as biological networks and collective behavior systems—remains challenging due to their intricate, heterogeneous structures. To address this, we propose GAUDI, the first unsupervised geometric deep learning framework for disentangling graph latent spaces. GAUDI introduces (1) a skip-connected encoder-decoder hourglass architecture integrating hierarchical graph pooling/upsampling with topology-preserving mechanisms; and (2) the first unsupervised explicit disentanglement of generative structural features from instance-specific noise in graphs, mapping isomorphic systems into a continuous, interpretable, structured latent space. Evaluated on small-world network modeling, super-resolution protein assembly analysis, Vicsek model representation, and human brain connectomic aging detection, GAUDI significantly improves graph classification accuracy and cross-scale interpretability, while uncovering domain-agnostic emergent structural principles.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Graphs provide a powerful framework for modeling complex systems, but their structural variability makes analysis and classification challenging. To address this, we introduce GAUDI (Graph Autoencoder Uncovering Descriptive Information), a novel unsupervised geometric deep learning framework that captures both local details and global structure. GAUDI employs an innovative hourglass architecture with hierarchical pooling and upsampling layers, linked through skip connections to preserve essential connectivity information throughout the encoding-decoding process. By mapping different realizations of a system - generated from the same underlying parameters - into a continuous, structured latent space, GAUDI disentangles invariant process-level features from stochastic noise. We demonstrate its power across multiple applications, including modeling small-world networks, characterizing protein assemblies from super-resolution microscopy, analyzing collective motion in the Vicsek model, and capturing age-related changes in brain connectivity. This approach not only improves the analysis of complex graphs but also provides new insights into emergent phenomena across diverse scientific domains.

Problem

• Analyzes and classifies complex graph structures with structural variability.
• Captures local and global graph features using unsupervised geometric deep learning.
• Disentangles invariant process-level features from stochastic noise in graphs.

Innovation

• Unsupervised geometric deep learning framework
• Hourglass architecture with hierarchical pooling
• Continuous structured latent space mapping