Metrics for network comparison using egonet feature distributions.

Area of Science:

• Network Science
• Graph Theory
• Data Analysis

Background:

• Quantifying network dissimilarity is crucial for tasks like identifying similar network ensembles or detecting changes in temporal network sequences.
• Existing methods often require network alignment or have high computational costs.

Purpose of the Study:

• To propose a novel, alignment-free method for quantifying network dissimilarity.
• To leverage local egonet features for a global network comparison approach.

Main Methods:

• A global network portrait is constructed by processing local node features, specifically degree, clustering coefficient, and egonet persistence.
• Dissimilarity measures are derived from the distributions of these egonet features across the network.
• Distribution functions are compared to compute network distances, without requiring network alignment.

Main Results:

• The proposed method demonstrates effectiveness in a classification test, accurately recognizing graphs from the same synthetic model.
• Performance is comparable to state-of-the-art graphlet-based methods.
• Computational requirements are similar to existing advanced techniques.

Conclusions:

• The proposed method offers a simple, flexible, and effective approach for network comparison.
• It provides a viable alternative to current network dissimilarity quantification techniques.