A hardware Markov chain algorithm realized in a single device for machine learning

• DOI: 10.1038/s41467-018-06644-w
• Publication Date: 2018-10-11T08:51:32Z
• Authors: He Tian; Xue-Feng Wang; Mohammad Ali Moha...; Guang-Yang Gou; Fan Wu; Yi Yang; Tian-Ling Ren
• Keywords: Reset (finance)

There is a growing need for developing machine learning applications. However, implementation of the algorithm consumes huge number transistors or memory devices on-chip. Developing capability in single device has so far remained elusive. Here, we build Markov chain based on native oxide two dimensional multilayer tin selenide. After probing electrical transport vertical oxide/tin selenide/tin heterostructures, sudden current jumps are observed during set and reset processes. Furthermore, five filament states observed. classifying into three chain, probabilities between each show convergence values after multiple testing cycles. Based this device, demo fixed-probability random generator within 5% error rate. This work sheds light as one hardware core with algorithm.