1: \begin{abstract}

2: 

3: Understanding causal relations is vital in scientific discovery. The process of

4: causal structure learning involves identifying causal graphs from observational

5: data to understand such relations. Usually, a central server performs this task,

6: but sharing data with the server poses privacy risks. Federated learning can

7: solve this problem, but existing solutions for federated causal structure

8: learning make unrealistic assumptions about data and lack convergence

9: guarantees. \tool\ is a federated constraint-based causal structure learning

10: scheme that learns causal graphs using a federated conditional independence

11: test, which examines conditional independence between two variables under a

12: condition set without collecting raw data from clients. \tool\ requires weaker

13: and more realistic assumptions about data and offers stronger resistance to data

14: variability among clients. \fedpc\ and \fedfci\ are the two variants of \tool\

15: for causal structure learning in causal sufficiency and causal insufficiency,

16: respectively. The study evaluates \tool\ using both synthetic datasets and

17: real-world data against existing solutions and finds it demonstrates encouraging

18: performance and strong resilience to data heterogeneity among clients.

19: 

20: \keywords{federated learning \and Bayesian network \and  probabilistic graphical model \and  causal discovery.}

21: \end{abstract}