题目：Transfer DAG Learning and Its Application to Biological Regulatory Networks

报告人：任明旸 副教授 （上海交通大学）

地点：致远楼101室

时间：2025年11月10日 16：00-17：30

摘要：Directed acyclic graph (DAG) is pivotal for modeling biological regulatory networks, such as brain functional connectivity networks and gene regulatory networks, yet data  from a single study is often limited for accurate DAG reconstruction. Although multi-source datasets provide augmented sample sizes, substantial heterogeneity persists across real-world cohorts. It remains an open question how to pool the heterogeneous data with only local similarity together for better DAG structure reconstruction in the target study. In this work, we introduce a novel set of structural similarity measures for DAGs and then present a transfer DAG learning framework by effectively leveraging information from auxiliary DAGs with node-level local similarities. Our theoretical analysis shows substantial improvement in terms of DAG reconstruction in the target study, which is in sharp contrast to most existing transfer learning methods requiring global similarity. The advantage of the proposed transfer DAG learning is also supported by extensive numerical experiments on both synthetic data and multi-source brain imaging and cancer genomics data.

All are welcome!