Circuit-based New Type-II Nonlinear Memristor for Neuromorphic Applications

Nonlinear memristors are more efficient than linear memristors for implementation in hardware-based neuromorphic applications. Very few researchers studied, the development of physical such nonlinear memristors in recent literature. Further, using this memristor real-life application is limited. Therefore, in this study to repeal its potential, a novel type II non-linear memristor is proposed. Using off-the-shelf components a simplified circuit-based type II non-linear memristor is developed through laboratory experiments. Some important characteristics are examined here. Finally, the proposed memristor-based crossbar array is utilized as a kernel of convolutional neural network (CNN) for application in handwritten pattern recognition. The simulation results depict 99.08% accuracy on the Modified National Institute of Standards and Technology (MNIST) handwritten digit dataset.