Applicability of adaptable reservoir computing to heterogeneous complex network structures

Determine whether the generalized adaptable reservoir computing scheme that substitutes failed oscillators using a single, small-size reservoir computer with state, coupling, and parameter-control input channels can be successfully applied to complex networks with heterogeneous structures so as to preserve the collective dynamics (such as the synchronization order parameter) over a sufficiently long period.

Background

The paper proposes a generalized adaptable reservoir computing (RC) approach that incorporates coupling signals and a parameter-control channel to substitute any oscillator in the globally coupled, heterogeneous Kuramoto model while sustaining the system’s synchronization order parameter for a time. The results demonstrate feasibility for globally coupled oscillators (homogeneous structural coupling).

However, real-world complex systems often feature heterogeneous network structures (e.g., nonuniform degree distributions and irregular topologies). The authors note that, while the scheme is expected to extend to networks with homogeneous structures, it is explicitly unknown whether it applies to networks with heterogeneous structures, leaving the broader generalization of the method unresolved.