报告题目：Artificial spiking neuron based on a single-photon avalanche diode and a VCSEL

报告人：V. N. Chizhevsky¹

1-B. I. Stepanov Institute of Physics, National Academy of Sciences of Belarus, 220072 Minsk, Belarus

Email: vnc@dragon.bas-net.by

报告时间：2023年8月26日上午9：00 – 10：00

报告地点：激光楼116（多功能厅）

报告邀请人：李念强教授

报告人简介：

Prof. V.N. Chizhevsky received the Ph.D. in laser physics from B.I. Stepanov Institute of Physics, National Academy of Sciences of Belarus, in 1992. His research interests include nonlinear laser dynamics and chaos, stochastic processes in the vertical surface-emitting laser, random number generation, ghost imaging, self-mixing interferometry etc.

报告摘要：

In recent years, considerable interest has been devoted to the development and study of artificial spiking neurons (ASN) with a high speed, a low power consumption, a scalability and a parallel operation. Such properties are necessary for the development of artificial spiking neural networks (SNN) of the third generation, which most closely mimic the operation of biological neural networks. Here an experimental realization and characterization of the artificial spiking neuron based on an optoelectronic pair “VCSEL - single photon avalanche diode” operating in few photon regime is presented. We show that basic properties of biological neurons, such as an existence of the threshold and the refractory period, the insensitivity to the effect of the stimuli strength above the threshold, and the dependence of the neuron fire rate of the stimuli strength, can be realized with such a type of artificial neuron. We also demonstrate an enhancement of propagation of weak subthreshold periodic signals in this ASN by the use of phenomenon of stochastic and vibrational resonance. Using this ASN in a stochastic mode of operation in combination with the software-implemented SNN of parallel-connected neurons, we show the possibility of the transmission of images through such the network practically without distortion.