Diffusive Memristors with Uniform and Tunable Relaxation Time for Spike Generation in Event-Based Pattern Recognition

摘要

A diffusive memristor is a promising building block for brain-inspired computinghardware. However, the randomness in the device relaxation dynamicslimits the wide-range adoption of diffusive memristors in large arrays. In thiswork, the device stack is engineered to achieve a much-improved uniformity inthe relaxation time (standard deviation σ reduced from ≈12 to ≈0.32 ms). Thememristor is further connected with a resistor or a capacitor and the relaxationtime is tuned between 1.13 μs and 1.25 ms, ranging from three orders ofmagnitude. The hierarchy of time surfaces (HOTS) algorithm, to utilize thetunable and uniform relaxation behavior for spike generation, is implemented.An accuracy of 77.3 is achieved in recognizing moving objects in the neuromorphicMNIST (N-MNIST) dataset. The work paves the way for buildingemerging neuromorphic computing hardware systems with ultralow powerconsumption.