Kim Tae-Hyeon, Lee Jaewoong, Kim Sungjoon, Park Jinwoo, Park Byung-Gook, Kim Hyungjin
Department of Electrical and Computer Engineering, Inter-University Semiconductor Research Center (ISRC), Seoul National University, Seoul 08826, Republic of Korea.
Department of Electronic Engineering, Inha University, Incheon 22212, Republic of Korea.
Nanotechnology. 2021 Apr 30;32(29). doi: 10.1088/1361-6528/abf0cc.
As interest in artificial intelligence (AI) and relevant hardware technologies has been developed rapidly, algorithms and network structures have become significantly complicated, causing serious power consumption issues because an enormous amount of computation is required. Neuromorphic computing, a hardware AI technology with memory devices, has emerged to solve this problem. For this application, multilevel operations of synaptic devices are important to imitate floating point weight values in software AI technologies. Furthermore, weight transfer methods to desired weight targets must be arranged for off-chip training. From this point of view, we fabricate 32 × 32 memristor crossbar array and verify the 3-bit multilevel operations. The programming accuracy is verified for 3-bit quantized levels by applying a reset-voltage-control programming scheme to the fabricated TiO/AlO-based memristor array. After that, a synapse composed of two differential memristors and a fully-connected neural network for modified national institute of standards and technology (MNIST) pattern recognition are constructed. The trained weights are post-training quantized in consideration of the 3-bit characteristics of the memristor. Finally, the effect of programming error on classification accuracy is verified based on the measured data, and we obtained 98.12% classification accuracy for MNIST data with the programming accuracy of 1.79% root-mean-square-error. These results imply that the proposed reset-voltage-control programming scheme can be utilized for a precise tuning, and expected to contribute for the development of a neuromorphic system capable of highly precise weight transfer.
随着对人工智能（AI）及相关硬件技术的兴趣迅速发展，算法和网络结构变得极为复杂，由于需要大量计算，导致了严重的功耗问题。神经形态计算作为一种带有存储器件的硬件AI技术应运而生，旨在解决这一问题。对于此应用，突触器件的多级操作对于模仿软件AI技术中的浮点权重值很重要。此外，必须安排向期望权重目标的权重转移方法用于片外训练。从这一角度出发，我们制造了32×32的忆阻器交叉阵列并验证了3位多级操作。通过对制造的基于TiO/AlO的忆阻器阵列应用复位电压控制编程方案，验证了3位量化级别的编程精度。之后，构建了由两个差分忆阻器组成的突触以及用于改进的美国国家标准与技术研究院（MNIST）模式识别的全连接神经网络。考虑到忆阻器的3位特性，对训练后的权重进行了训练后量化。最后，基于测量数据验证了编程误差对分类精度的影响，对于MNIST数据，我们在编程精度为均方根误差1.79%的情况下获得了98.12%的分类精度。这些结果表明，所提出的复位电压控制编程方案可用于精确调谐，并有望为能够进行高精度权重转移的神经形态系统的发展做出贡献。
