Hfo2-Based Bipolar Resistive Memory Using 2T1M Synapse for Spike Time Dependant Plasticity
Abstract
Metal oxide-based resistive (or memristive) switching devices are being used in storage, arithmetic, and neuromorphic computing systems. Neuro-inspired computing requires a high density of active connections (synapses) with minimal power consumption, and these devices meet those requirements with their tiny dimensions, low power consumption, and create high efficiency. A new artificial synapse system is presented in this paper, which consists of a memristive switch linked to two transistors that gate learning and interaction processes. The STDP is performed by the proper forming of presynaptic and post synaptic spikes. Stochastic STDP is observed in integrated artificial synapse studies due to the intrinsic unpredictability of the nano-size switch set/reset method and the switch's different responses to a given stimulus depending on the starting condition. Memory switching simulations using different models confirm the experimental findings. A neural 2-layer network at the system level and a simple STDP model are simulated.