Metaplasticity
is one of the synaptic functions in the brain, known
as the plasticity of synapse plasticity. It allows some specific functions
of the brain, such as continual learning, rapid homeostasis, and synaptic
tag-and-capture functions. Here, we propose an electrically triggerable
HfAlO
x
-based metaplastic device that exhibits
reversible multimode synaptic behaviors. The origin of the synaptic
characteristic change is the change in the serial and parallel resistance
components of the memristive system, accompanied by unbalanced oxygen
flux during the set and reset switching, triggered by the priming
activity. Four distinguishable potentiation and depression characteristics
(two stable modes and two transition modes) are suggested, which can
be explained using a plausible schematic model. A hardware-based adaptive
learning system for artificial neural networks is proposed as a potential
application of the metaplastic device.