Fig. 7: Ordered (λ < 0), edge-of-chaos (λ ≈ 0) and chaotic (λ > 0) network states.
From: Avalanches and edge-of-chaos learning in neuromorphic nanowire networks
Each column corresponds to a single NWN simulation for a triangular AC input (n = 20) cycles, with A = 1.25 V and f = 0.1, 0.5, 0.85 Hz (left to right). Top row: I − V curves, with arrows indicating direction of voltage control. Bottom row: corresponding network conductance, Gnw. First column (a, d): f = 0.1 Hz, r = 7.7 × 103 and λ = − 2.6 s−1; all junction filament states return to 0 between cycles (reaching V = 0), resulting in symmetric repeatable I − V and Gnw − V cycles. Second column (b, e): states near the edge-of-chaos, with f = 0.5 Hz, r = 7.1 × 103 and λ = 0.4 s−1; network does not fully deactivate as polarity of voltage is reversed. Third column (c, f): f = 0.85 Hz, r = 5.9 × 103 and λ = 4.1 s−1; network trajectories are chaotic. A 100 nanowire, 261 junction network is used, but qualitatively similar results are found on a range of network sizes and densities.
