Researchers Propose a Framework for Modern Analog Computing
A review by researchers from Peking University, IBM Zurich Research Center and Unconventional AI proposes a definition of modern analog computing centered on matrix operations using devices such as resistive random-access memory (RRAM). The review identifies differential-equation solving, matrix-equation solving and matrix-vector multiplication as core primitives. In related work reported by Annax Zhixin’s team, an analog matrix-computing circuit achieved a closed-loop solve in under 100 nanoseconds, 24-bit fixed-point precision and scaling to 512×512 matrices.
The proposed framework seeks to unify scattered analog-computing techniques and sets out primitives, hardware requirements and a path toward higher precision and scalability. The report also identifies embodied-AI control, 6G massive-MIMO processing and large-model optimization as application areas or development priorities, including demonstrations and chip plans.
模拟计算进入矩阵时代!北大联手硅谷IBM,写下新定义 – 量子位
< img id="wx_img" src="https://www.qbitai.com/wp-content/uploads/imgs/qbitai-logo-1.png" width="400" height="400">
模拟计算进入矩阵时代!北大联手硅谷IBM,写下新定义

克雷西 2026-11-03 11:16:23 来源:量子位
GPU迭代一百次的事,它一步做完
#Peking University#IBM Zurich Research Center#Unconventional AI#Nature Electronics#RRAM#NVIDIA H100#5G-A#6G#Annax Zhixin