FALCON or how to compute measures time efficiently on dynamically evolving dense complex networks?

Area of Science:

• Complex network analysis in biology, medicine, neuroscience, psychology, and sociology.
• Focus on biological networks, including functional spatiotemporal brain activations and changes due to neuropsychiatric pathologies.

Background:

• Analyzing complex networks involves calculating meaningful measures, which can be time-consuming for large or numerous networks.
• Existing libraries are often optimized for large, sparse networks, posing challenges for analyzing many smaller, dense networks, common in fields like brain imaging.

Purpose of the Study:

• To introduce FALCON, a novel library specifically designed for the efficient exploration of dense complex networks.
• To provide optimized computational methods for analyzing large numbers of smaller, dense networks.

Main Methods:

• FALCON utilizes a multi-core approach with extensive code and hardware optimizations.
• It offers an alternative massively parallel GPU implementation for computationally intensive measures.
• The library supports 12 different measures for undirected-unweighted, undirected-weighted, and directed-unweighted networks.

Main Results:

• FALCON demonstrates efficient analysis of dense complex networks, significantly reducing calculation times compared to standard methods.
• The multi-core and GPU implementations provide substantial performance gains.
• An integrated benchmark aids users in selecting the most appropriate library for their specific network analysis needs.

Conclusions:

• FALCON offers a powerful and efficient solution for analyzing dense complex networks, particularly valuable in research involving numerous smaller networks.
• The library's optimized implementations and benchmarking tool enhance the feasibility and accuracy of complex network analysis in various scientific domains.