DAG Learning
DAG learning focuses on inferring the structure of directed acyclic graphs (DAGs) from data, representing causal relationships between variables. Current research emphasizes developing efficient and scalable algorithms, including those based on continuous optimization, neural networks (like neural spacetimes), and Bayesian methods, to overcome the computational challenges posed by the combinatorial nature of the problem. These advancements are improving the accuracy and applicability of causal discovery across diverse fields, from biological network analysis and traffic flow prediction to personalized recommendations and multi-agent systems. The resulting DAGs provide valuable insights into complex systems and enable more robust and interpretable models.
Papers
MM-DAG: Multi-task DAG Learning for Multi-modal Data -- with Application for Traffic Congestion Analysis
Tian Lan, Ziyue Li, Zhishuai Li, Lei Bai, Man Li, Fugee Tsung, Wolfgang Ketter, Rui Zhao, Chen Zhang
Discovering Dynamic Causal Space for DAG Structure Learning
Fangfu Liu, Wenchang Ma, An Zhang, Xiang Wang, Yueqi Duan, Tat-Seng Chua