Causality-Driven Disentangled Representation Learning in Multiplex Graphs

2026-03-25

AI Summary

This work addresses the entanglement of shared and layer-specific information in multilayer graphs, which hinders the generalizability and interpretability of learned representations. To resolve this, we propose the first causality-based self-supervised framework that rigorously disentangles shared and private representations via backdoor adjustment. Specifically, our approach aligns cross-layer shared embeddings to capture global structural patterns while enhancing private embeddings to preserve layer-specific signals. Evaluated on both synthetic and real-world multilayer graph datasets, the method significantly outperforms existing baselines, demonstrating improved model robustness and interpretability. Notably, this study represents the first successful integration of causal inference into the task of disentangling representations in multilayer graphs.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

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
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Learning representations from multiplex graphs, i.e., multi-layer networks where nodes interact through multiple relation types, is challenging due to the entanglement of shared (common) and layer-specific (private) information, which limits generalization and interpretability. In this work, we introduce a causal inference-based framework that disentangles common and private components in a self-supervised manner. CaDeM jointly (i) aligns shared embeddings across layers, (ii) enforces private embeddings to capture layer-specific signals, and (iii) applies backdoor adjustment to ensure that the common embeddings capture only global information while being separated from the private representations. Experiments on synthetic and real-world datasets demonstrate consistent improvements over existing baselines, highlighting the effectiveness of our approach for robust and interpretable multiplex graph representation learning.

Problem

• multiplex graphs
• disentangled representation
• causality
• common and private information
• representation learning

Innovation

• causal inference
• disentangled representation
• multiplex graphs
• backdoor adjustment
• self-supervised learning