JSAP Spring / Autumn Meeting

With the rapid development and emergent wild applications of artificial intelligence software, to achieve artificial intelligence hardware has been optimistic recently. One main topic is to mimic the brain’s memorization functions on artificial synapses—materials have memristive properties^1-4. This work we studied learning and forgetting relationship on Ag@TiO₂ nanowire random network. Unlike common understanding of conventional resistance-based random access memory, here in Ag@TiO₂ network, connectivity in network plays the key role on reactivation of a memory. After activation of a memory, resistance usually rapid increases to a non-conductive state in the rest period. The reactivation time of the memory grows with rest period time (see Figure). This phenomena is more close to the brain’s behavior when handling multi-tasks and could help to improve performance on neuromorphic computation.