A mean-field approximation-based linearization framework for network reconstruction from binary time series.

Area of Science:

• Network science
• Computational biology
• Statistical physics

Background:

• Reconstructing complex networks from limited binary-state time series data presents a significant challenge.
• Existing methods often rely on known dynamical rules or empirically determined linear equations, limiting generality and interpretability.

Purpose of the Study:

• To develop a novel, broadly applicable, and interpretable network reconstruction method for binary-state time series data.
• To address limitations of existing approaches, particularly their dependence on specific dynamical rules or parameter sensitivity.

Main Methods:

• Proposes a network reconstruction method based on linearization grounded in mean-field approximation.
• Exploits the common feature of binary-state dynamics where node activation depends on active neighbors.
• Develops a non-blocking, parameter-free alternative to computationally complex blocking strategies.

Main Results:

• The proposed method enhances interpretability through mean-field approximation.
• Demonstrates theoretical and empirical evidence of reconstruction performance comparable to ideal blocking methods.
• Verifies effectiveness and robustness on artificial and real networks using noisy data.

Conclusions:

• The developed method provides a general and robust approach for network reconstruction from binary-state time series.
• Mean-field approximation improves interpretability and applicability across diverse network dynamics.
• The parameter-free, non-blocking strategy offers computational advantages without sacrificing performance.