Organic memcapacitor-based neural network for diverse signal recognition

Memcapacitors emerge as a promising electronic element for low-power artificial neural networks, whereas research on their applications remains limited due to the challenge of using capacitance-based state variables. We report the probing into transient responses of an organic memcapacitor and how the transient features endow the device with the function of multiply-and-accumulate operation. The neural network based on the organic memcapacitor array is capable of processing spatial and temporal signal recognition, as well demonstrated by its excellent performance in fashion image recognition and arrhythmia detection, achieving the classification accuracies of 87 % and 99 %, respectively. This work initiates a potential approach to adopting dynamic characteristics, rather than steady-state behaviors, of memcapacitors for implementing neuromorphic computing tasks.