Spectral coarse graining and rescaling for preserving structural and dynamical properties in graphs

Area of Science:

• Computational neuroscience
• Graph theory
• Network science

Background:

• Analyzing large-scale complex networks, such as human brain activity, presents significant computational challenges.
• Existing methods may struggle to retain crucial dynamic and topological information during data reduction.

Purpose of the Study:

• To introduce a novel graph renormalization procedure for creating multi-scale, reduced-complexity graph representations.
• To demonstrate the method's ability to preserve essential network dynamics and topological structures.
• To apply this method to electroencephalogram (EEG) data for analyzing brain activity.

Main Methods:

• Developed a graph renormalization technique utilizing the coarse-grained Laplacian.
• Applied the method to graphs constructed from human electroencephalogram (EEG) recordings.
• Analyzed the reduced-complexity representations to identify emergent behaviors and dynamic reorganizations.

Main Results:

• The renormalization procedure effectively reduces graph complexity while preserving dynamics and large-scale topology.
• Analysis of EEG data revealed coordinated neuronal activity patterns indicative of collective behavior.
• Observed dynamic reorganization of brain activity across scales, with distinct patterns during rest (generalized) and attention (specialized, scale-invariant in the occipital lobe).

Conclusions:

• The coarse-grained Laplacian-based renormalization offers an effective approach for analyzing large, complex graph data, particularly in neuroscience.
• The method facilitates the discovery of emergent collective behaviors and dynamic brain network reorganizations.
• This technique provides valuable insights into brain function across different cognitive states and scales.