Sublinear evaluation of complex networks for extensive exploration of configurations for critical scenarios and decision making

Artificial Intelligence in Complex Networks has contributed to several relevant fields involving energy systems, computer networks, environment, agriculture, health, and social organizations. However, investigations concerning multiple heterogeneous networks have been less frequent. Mixed systems usually require fine-grained data to retain a sufficient amount of details from each network. This type of modeling may enable the investigation of emerging behaviors or synergies. For example, a decision making may require the search for an improved network configuration (involving coarse and fine modifications on devices, procedures, and settings) with the lowest possible cost to soon mitigate effects from climate changes or other types of "attacks" (from economic crises, calamities, and recent pandemics). The generation of robust configurations for heterogeneous networks involves some challenges, pointed out in this paper. Among them, the efficient calculus of load flows has been one of the main challenges. To overcome it, we propose a load flow algorithm with sublinear time complexity for the construction and evaluation of several configurations. The new algorithm scales well and can deal with nonlinear dynamics in evaluations of entire sets of fine-grained network models that it may involve.