Identification of network interactions from time series data: An iterative approach

Area of Science:

• Complex Systems Science
• Network Science
• Data Analysis

Background:

• Understanding complex networks requires determining connectivity from time series data.
• High-dimensional networks often have limited available data.
• Existing methods face challenges with scalability and noise.

Purpose of the Study:

• To develop a robust and scalable method for network inference from time series data.
• To address limitations of current approaches in handling high-dimensional and noisy data.
• To improve the accuracy of network structure determination.

Main Methods:

• Formulated network inference as a bilinear optimization problem.
• Developed an iterative algorithm with sequential initialization.
• Tested the approach on diverse simulated and experimental datasets.

Main Results:

• Demonstrated scalability with increasing network size.
• Showcased robustness against measurement noise and parameter variations.
• Achieved superior inference accuracy compared to existing methods across various dynamics (oscillatory, non-oscillatory, chaotic).

Conclusions:

• The proposed iterative algorithm offers a powerful and accurate solution for network inference.
• The method is reliable for complex, high-dimensional systems with limited data.
• This technique advances the understanding of network connectivity in diverse scientific fields.