Design of Discrete Hopfield Neural Network Using a Single Flux Quantum Circuit
Design of Discrete Hopfield Neural Network Using a Single Flux Quantum Circuit
复制标题
使用单通量量子电路的离散 Hopfield 神经网络设计
DOI：
--
复制
发表时间：
2022
影响因子：
1.8
通讯作者：
N. Yoshikawa
中科院分区：
文献类型：
--
作者：
Houwen He;Y. Yamanashi;N. Yoshikawa
超导单通量量子（SFQ）逻辑家族以其高开关速度和低功耗的优势，被认为是解决后摩尔时代能耗危机的最佳选择。在机器学习领域，技术和计算需求正在快速增长（例如，图像识别和自然语言处理），SFQ电路的实现具有很大的潜力。在这项研究中，我们研究并实现了离散Hopfield神经网络（DHNN）使用SFQ电路。DHNN是一种二进制神经网络，它比标准的全精度神经网络具有更少的信息;它还提供了更高的处理速度。它主要用于模式识别和恢复。我们设计了两种模式的DHNN电路，每种模式有八个元件。该电路工作在超过50 GHz的时钟频率。
The superconductor single flux quantum (SFQ) logic family has been recognized as a promising candidate to resolve the energy consumption crisis in the post-Moore era, owing to its high switching speed and low power consumption. In the field of machine learning, where technology and computational requirements are growing rapidly (e.g., image recognition and natural language processing), there is great potential for the implementation of SFQ circuits. In this study, we investigate and implement a discrete Hopfield neural network (DHNN) using SFQ circuits. A DHNN is a binary neural network with less information than a standard full precision neural network; it also provides a higher processing speed. It is mainly used for pattern recognition and recovery. We designed the DHNN circuit with two patterns, each with eight elements. The circuit operates at the clock frequency of more than 50 GHz.