Generating realistic scaled complex networks.

Area of Science:

• Network Science
• Computational Science

Background:

• Generative models are crucial for understanding network science, enabling the study of statistical patterns and the development of computational methods.
• Existing models aim for realistic network generation but often face limitations in comprehensive realism and scalability.

Purpose of the Study:

• Introduce a novel generative model, ReCoN (Replicating and Constructing Networks).
• Evaluate ReCoN's ability to generate structurally similar replicas of original networks.
• Assess ReCoN's capability for significant network size scaling.
• Compare ReCoN's performance against state-of-the-art network generation methods.

Main Methods:

• Development of the ReCoN generative model.
• Comparative experimental study involving fitting models to original networks.
• Evaluation of structural similarity and property preservation at micro- and macroscopic scales.
• Analysis of ReCoN's performance in scaling network data.

Main Results:

• ReCoN demonstrates superior performance compared to many existing state-of-the-art network generation methods.
• ReCoN effectively models given networks, preserving key micro- and macroscopic properties.
• ReCoN successfully scales exemplar network data by orders of magnitude.

Conclusions:

• ReCoN is a scalable and effective tool for network modeling and data augmentation.
• The model offers a promising direction for advancing generative approaches in network science.
• Future research can explore open problems and further enhance generative model capabilities.