Long-time-constant leaky-integrating oxygen-vacancy drift-diffusion FET for human-interactive spiking reservoir computing
Hisashi Inoue, Hiroto Tamura, Ai Kitoh, Xiangyu Chen, Zolboo Byambadorj, Takeaki Yajima, Yasushi Hotta, Tetsuya Iizuka, Gouhei Tanaka, Isao H. Inoue · 2023
Reservoir computing is one of the best feasible means of information processing that does not require substantial computation, such as backpropagation. Implementation using spiking neural networks is promising for real-time and low-energy computation that can be completed by edge devices alone. The key parameter is a time constant $\tau_{\mathrm{n}}$ of the neurons, which should be close to that of external stimuli and can be on the order of milliseconds for human-interactive reservoir computing. Here, we present a slow spiking neuron based on an oxygen-vacancy drift-diffusion field-effect transistor (ODD-FET). In these transistors, the slow migration of oxygen vacancies mimics a leaky integration with $\tau_{\mathrm{n}}$ on the order of 10 to 100 ms. The feasibility of the proposed ODDFET on the human-interactive application is demonstrated by a spiking reservoir composed of these neurons, which can accurately detect handwriting anomalies.