Coarse-to-Fine Learning of Dynamic Causal Structures

2026-02-25

AI Summary

This work addresses the challenge of learning fully dynamic, time-varying causal structures in real-world systems, overcoming the limitations of traditional methods that rely on stationarity assumptions. The authors propose DyCausal, a novel framework featuring a coarse-to-fine learning strategy: it first captures causal patterns within coarse-grained time windows using convolutional networks and then refines the estimates to each individual time step via linear interpolation. To ensure the validity of the inferred causal graphs, DyCausal incorporates an acyclicity constraint based on matrix norm scaling. Extensive experiments demonstrate that DyCausal significantly outperforms existing approaches on both synthetic and real-world datasets, achieving efficient, stable, and stationarity-free discovery of fully dynamic causal structures.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Learning the dynamic causal structure of time series is a challenging problem. Most existing approaches rely on distributional or structural invariance to uncover underlying causal dynamics, assuming stationary or partially stationary causality. However, these assumptions often conflict with the complex, time-varying causal relationships observed in real-world systems. This motivates the need for methods that address fully dynamic causality, where both instantaneous and lagged dependencies evolve over time. Such a setting poses significant challenges for the efficiency and stability of causal discovery. To address these challenges, we introduce DyCausal, a dynamic causal structure learning framework. DyCausal leverages convolutional networks to capture causal patterns within coarse-grained time windows, and then applies linear interpolation to refine causal structures at each time step, thereby recovering fine-grained and time-varying causal graphs. In addition, we propose an acyclic constraint based on matrix norm scaling, which improves efficiency while effectively constraining loops in evolving causal structures. Comprehensive evaluations on both synthetic and real-world datasets demonstrate that DyCausal achieves superior performance compared to existing methods, offering a stable and efficient approach for identifying fully dynamic causal structures from coarse to fine.

Problem

• dynamic causal structure
• time-varying causality
• causal discovery
• non-stationary time series
• fine-grained causal graphs

Innovation

• dynamic causal discovery
• coarse-to-fine learning
• time-varying causal graphs
• acyclicity constraint
• convolutional causal modeling