论文部分内容阅读
突触可塑性为神经网络的学习机制提供了基础。基于单个半导体光放大器(SOA)的非线性偏振旋转(NPR)和交叉增益调制(XGM)效应实现了反脉冲时间依赖可塑性(anti-STDP)学习机制。通过调整SOA驱动电流,可以实现长时程增强窗口(LTP)和长时程抑制窗口(LTD)的高度和宽度调整,能更好地模拟神经网络。实验测量得到的anti-STDP曲线与生物系统中测量得到的学习曲线相吻合。使用该anti-STDP光路得到的学习曲线的时间窗口约为几百皮秒,其速度是人类大脑STDP学习机制的108倍。由于该anti-STDP光路系统简单,且SOA易于与其他器件集成,该anti-STDP光路可以用于实现大规模超快神经拟态计算系统。
Synaptic plasticity provides the basis for the learning mechanism of neural networks. The anti-pulse time-dependent plasticity (anti-STDP) learning mechanism is based on the nonlinear polarization rotation (NPR) and cross-gain modulation (XGM) effects of a single semiconductor optical amplifier (SOA). By adjusting the SOA drive current, height and width adjustments can be realized for long-term enhanced window (LTP) and long-term suppressed window (LTD) for better modeling of neural networks. The anti-STDP curves obtained from the experimental measurements are in good agreement with the measured learning curves in the biological system. The learning curve obtained with this anti-STDP optical path has a time window of about a few hundred picoseconds, which is 108 times faster than the human brain’s STDP learning mechanism. Due to the simplicity of the anti-STDP optical system and its ease of integration with other devices, the anti-STDP optical path can be used to implement large-scale, ultra-fast neuromorphic computing systems.