Opt Lett. 2023 Apr 15;48(8):2150-2153. doi: 10.1364/OL.485215.
Hopfield networks are iterative procedures able to solve combinatorial optimization problems. New studies regarding algorithm-architecture adequacy are fostered by the re-emergence of hardware implementations of such methods in the form of Ising machines. In this work, we propose an optoelectronic architecture suitable for fast processing and low energy consumption. We show that our approach allows effective optimization relevant to statistical image denoising.
Hopfield 网络是一种迭代过程，能够解决组合优化问题。由于以 Ising 机形式出现的此类方法的硬件实现的重新出现，关于算法-体系结构充分性的新研究得到了促进。在这项工作中，我们提出了一种适合快速处理和低能耗的光电体系结构。我们表明，我们的方法允许对统计图像去噪进行有效的优化。
