A novel constraint-based structure learning algorithm using marginal causal prior knowledge.

Area of Science:

• Causal inference
• Machine learning
• Network analysis

Background:

• Causal discovery aims to infer causal relationships from data.
• Incorporating prior knowledge enhances causal discovery performance.
• Marginal causal relations represent existing directed paths in causal models.

Purpose of the Study:

• To develop a method for integrating marginal causal relations into causal discovery.
• To improve the effectiveness and efficiency of structure learning algorithms.

Main Methods:

• Propose the Marginal Prior Causal Knowledge PC (MPPC) algorithm.
• Develop theorems for conditional independence properties using observational data and marginal causal relations.
• Compare MPPC with existing structure learning methods via simulations and real-world networks.

Main Results:

• The MPPC algorithm successfully incorporates marginal causal relations.
• MPPC demonstrates superior effectiveness and efficiency compared to other constraint-based structure learning methods.
• Validation through simulation studies and analysis of real-world networks.

Conclusions:

• MPPC offers an effective approach for causal discovery with prior knowledge.
• The algorithm enhances causal structure learning by leveraging marginal causal relations.
• MPPC provides a more efficient and accurate method for causal inference.