[8p-C18-1~8] 13.5 Semiconductor devices and related technologies

[8p-C18-1] Spike-Timing-Dependent synaptic Plasticity (STDP) of metal-oxide devices for brain-inspired learning LSIs

〇Takao Marukame¹, Reika Ichihara¹, Kumiko Nomura¹, Shin-ichi Yasuda¹, Yuichiro Mitani¹ (1.Toshiba corp.)

Keywords:Memristor, Spike-Timing-Dependent synaptic Plasticity (STDP), Deep learning

Neuromorphic devices have attracted much attention for artificial intelligence applications, such as image recognition and speech recognition with great performance exceeding human beings, based on deep artificial neural networks. Spike-timing-dependent plasticity (STDP) is known as a mechanism for updating memory during learning in our biological neural networks. We observed STDP in metal-oxide (AlO_x/TiO_x) memristive devices having CMOS compatibility and then established the STDP protocols that can control analog resistive changes periodically with long range stabilities.