Interpretable Graph Learning Over Sets of Temporally-Sparse Data

2025-05-25

AI Summary

Modeling asynchronous, sparse, and multi-source temporal signals—common in healthcare and other domains—remains challenging due to irregular sampling and complex dynamic dependencies across heterogeneous sources. Method: This paper proposes the Graph Mixed Additive Network (GMAN), a novel “explainable-by-design” architecture that jointly learns time-aware adjacency structures and employs differentiable graph mixing to directly capture evolving inter-signal dependencies. GMAN supports multi-granularity attribution at node-, graph-, and subset-levels. Contribution/Results: We theoretically establish GMAN’s strong expressive power and, for the first time, demonstrate its unified applicability across disparate domains—specifically in-hospital mortality prediction and fake news detection. On real-world clinical data, GMAN achieves a 4-percentage-point AUROC improvement over state-of-the-art baselines. It transfers robustly to fake news detection without architectural modification. Attribution analysis uncovers clinically meaningful transition patterns during critical care episodes, enhancing interpretability for high-stakes decision-making.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches 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: Generative AI / large language models and their impact on social systems

Abstract

Real-world medical data often includes measurements from multiple signals that are collected at irregular and asynchronous time intervals. For example, different types of blood tests can be measured at different times and frequencies, resulting in fragmented and unevenly scattered temporal data. Similar issues of irregular sampling of different attributes occur in other domains, such as monitoring of large systems using event log files or the spread of fake news on social networks. Effectively learning from such data requires models that can handle sets of temporally sparse and heterogeneous signals. In this paper, we propose Graph Mixing Additive Networks (GMAN), a novel and interpretable-by-design model for learning over irregular sets of temporal signals. Our method achieves state-of-the-art performance in real-world medical tasks, including a 4-point increase in the AUROC score of in-hospital mortality prediction, compared to existing methods. We further showcase GMAN's flexibility by applying it to a fake news detection task. We demonstrate how its interpretability capabilities, including node-level, graph-level, and subset-level importance, allow for transition phases detection and gaining medical insights with real-world high-stakes implications. Finally, we provide theoretical insights on GMAN expressive power.

Problem

• Handling irregular and asynchronous temporal medical data
• Learning from sparse heterogeneous signals across domains
• Providing interpretable insights for high-stakes decision-making

Innovation

• Graph Mixing Additive Networks for sparse data
• Handles irregular temporal signals effectively
• Interpretable node and graph-level importance analysis