Neuromorphic Atomic Switch Networks

摘要: Efforts to emulate the formidable information processing capabilities of brain through neuromorphic engineering have been bolstered by recent progress in fabrication nonlinear, nanoscale circuit elements that exhibit synapse-like operational characteristics. However, conventional techniques are unable efficiently generate structures with highly complex interconnectivity found biological neuronal networks. Here we demonstrate physical realization a self-assembled device which implements basic concepts systems neuroscience hardware-based platform comprised over billion interconnected atomic-switch inorganic synapses embedded network silver nanowires. Observations activation and passive harmonic generation collective response input stimulus agreement theoretical predictions. Further, emergent behaviors unique atomic switches akin function observed, namely spatially distributed memory, recurrent dynamics feedforward subnetworks. These devices display functional characteristics required for implementing unconventional, biologically neurally inspired computational methodologies synthetic experimental system.