Evolutionary optimization of network reconstruction from derivative-variable correlations

Abstract

Topologies of real-world complex networks are rarely accessible, but can often be reconstructed from experimentally obtained time series via suitable network reconstruction methods. Extending our earlier work on methods based on statistics of derivative-variable correlations, we here present a new method built on integrating an evolutionary optimization algorithm into the derivative-variable correlation method. Results obtained from our modi cation of the method in general outperform the original results, demonstrating the suitability of evolutionary optimization logic in network reconstruction problems. We show the method's usefulness in realistic scenarios where the reconstruction precision can be limited by the nature of the time series. We also discuss important limitations coming from various dynamical regimes that time series can belong to.

This work was founded by the EU via H2020 Marie SklodowskaCurie project COSMOS, grant no. 642563. R G A acknowledges funding from the Volkswagen foundation, the Spanish Ministry of Economy and Competitiveness (Grant FIS2014-54177-R) and the CERCA Programme of the Generalitat de Catalunya. Z L acknowledges funding from the Slovenian Research Agency via program Complex Networks P1-0383 and project J5- 8236.