Detectability of macroscopic structures in directed asymmetric stochastic block model.

Area of Science:

• Network science
• Statistical physics

Background:

• Identifying macroscopic structures is crucial for understanding complex networks.
• The stochastic block model is a standard tool for network analysis.
• The effect of directed links on structure detectability is not fully understood.

Purpose of the Study:

• To investigate the impact of link directionality on the detectability of macroscopic network structures.
• To develop a framework for analyzing directed networks within the stochastic block model.
• To characterize the phase transition in directed network structure detection.

Main Methods:

• Construction of a class of nontrivially detectable directed networks based on the stochastic block model.
• Application of the belief propagation method to derive closed-form solutions.
• Numerical simulations to identify distinct phases of detection.

Main Results:

• The belief propagation method yields closed-form solutions for directed networks.
• The detectability transition line is dependent on network assortativity and asymmetry.
• A 'hard phase' for structure detection exists near the transition point in directed networks.

Conclusions:

• Link directionality significantly alters the detectability of macroscopic network structures.
• Assortativity and asymmetry are key factors governing the phase transition in directed networks.
• The identified 'hard phase' presents challenges for community detection in directed networks.