⚡ 本页包含 AI 生成的分析内容，仅供参考
本文提出了一款512Gb的存内计算3D NAND闪存，支持相似向量匹配操作，适用于边缘AI设备。该设计在3D NAND中直接实现计算，解决了传统架构中数据移动的瓶颈，适用于人脸搜索和音频纹理等机器学习任务。
Bo-Rong Lin1, Huai-Mu Wang1, Yen-Po Lin1, Yu-Chao Lin1, Chih-Chang Hsieh1, Chia-Ming Hu1, Yi-Ting Lai1, Han-Sung Chen1, Yuan-Hao Chang4, Hsiang-Pang Li1, Tei-Wei Kuo3,5, Keh-Chung Wang1, Meng-Fan Chang2, Chun-Hsiung Hung1, Chih-Yuan Lu1 Macronix, Hsinchu, Taiwan National Tsing Hua University, Hsinchu, Taiwan 3 National Taiwan University, Taipei, Taiwan 4 Academia Sinica, Taipei, Taiwan 5 City University of Hong Kong, Hong Kong, China 1 2 Similar-vector-matching (SVM) applications for unstructured vectors that are generated via machine-learning methods, such as face search and audio texturing from a dataset for access control systems, are frequently operated on edge devices, as depicted in Fig. 7.5.1. The SVM operation [1-3] typically comprises of (1) in the offline phase, the → and stored extracted raw vectors (VRAW) are obtained from machine learning approaches → in non-volatile NAND Flash; (2)
Han-Wen Hu1,2, Wei-Chen Wang1,3, Chung-Kuang Chen1, Yung-Chun Lee1,