We present a method that reconstructs both the links and their relative coupling strength of bidirectional weighted networks. Our method requires only measurements of node dynamics as input. Using several examples, we demonstrate that our method can give accurate results for weighted random and weighted scale-free networks with both linear and nonlinear dynamics.
DOI:https://doi.org/10.1103/PhysRevE.91.030801
©2015 American Physical Society
versus for (a) WR1 with Rössler and consensus (inset) dynamics, (b) WR2 with logistic dynamics (inset shows versus with solid line for the theoretical result [Eq. (6)]), (c) WSF1 with FHN1 and Rössler (inset) dynamics, (d) WSF2 with FHN2 and consensus (inset) dynamics, (e) WR2 with cubic coupling, and (f) WR1 with FHN3 dynamics for and (inset). In all the other cases, . Dashed line is .
Reuse & PermissionsComparison of the reconstructed with the actual for (a) WR1 with Rössler dynamics, (b) WSF1 with FHN1 dynamics, (c) WSF2 with FHN2 dynamics, and (d) WSF2 with Rössler dynamics. for (a) and (d) and for (b) and (c). Undetected links have but while links with but are incorrectly reconstructed. The average absolute error of nonzero is less than 7% for (a), less than and about 20% for (b) and (c), and about 10% for (d).
Reuse & PermissionsComparison of of the reconstructed for consensus (circles), Rössler (triangles), FHN1 (squares), and FHN2 (diamonds) dynamics with against the actual distribution (solid line) for (a) WR1 and WR2, (b) WSF1, and (c) WSF2 networks. Dashed line in (b) and (c) is the result for FHN2 dynamics with .
Reuse & PermissionsComparison of the reconstructed (circles) and the actual for (a) WR1 with Rössler dynamics, (b) WSF1 with FHN1 dynamics, (c) WSF2 with FHN2 dynamics, and (d) WSF2 with Rössler dynamics. In (b), we show also (triangles).
Reuse & Permissions
