[19p-233-1~12] Create a path of future semiconductor devices by new materials and processes

[19p-233-3] STDP characterization of MgO-based synapse devices and its application to neural networks

〇Takao Marukame¹, Radu Berdan¹, Kumiko Nomura¹, Yoshifumi Nishi¹ (1.Toshiba Corp. RDC)

Keywords:neural network, STDP, neuromorphic hardware

In the field of artificial intelligence, multilayer neural networks and deep learning have recently attracted much attention, and the related brain-inspired hardware has been intensively studied. Spike-timing dependent plasticity (STDP) is known as a principal of synapses in brain, and is targeted to realize its mimic using solid state devices. We have reported AlO_x/TiO_x devices showing STDP characteristics. In this work, we observed STDP on MgO-based devices and considered its applications for neural network hardware.