Fig. 4: Efficiency of Adjoint Optimization. | Nature Communications

Fig. 4: Efficiency of Adjoint Optimization.

From: In-situ physical adjoint computing in multiple-scattering electromagnetic environments for wave control

Fig. 4

Efficiency of the Adjoint Optimization as a function of the number of controlled parameters and network complexity htop for: a TMT modality; b CPA modality; and (c) signal invisibility. Solid lines correspond to three representative networks with htop ≈ 1 (blue), 2 (orange), 3 (green). Each network has V = 75 vertices with Vbulk = 50 bulk (interior) vertices whose connectivity is varied to tune htop, and VTL = 25 boundary vertices used to attach transmission lines (TLs). Each TL vertex connects to the bulk via two bonds (valency vTL = 2), which keeps the TL-network coupling constant fixed as the bulk connectivity changes (see Methods). In all cases, performance improves as the fraction σ of tunable bulk-bulk bonds (controlled parameters) increases. For h ≈ 1, 2, 3 the total number of bulk bonds are 73, 193, 529, respectively.

Back to article page