← 返回论文列表 📄 下载原文 PDF  ISSCC 2023 · 7.9
ISSCC 2023Session 7 · SRAM COMPUTE-IN-MEMORYAI / ML

CTLE-Ising: A 1440-Spin Continuous-Time Latch-Based Ising Machine with One-Shot Fully-Parallel Spin Updates Featuring Equalization of Spin States

⚡ 本页包含 AI 生成的分析内容,仅供参考

📋 论文概要

该论文提出了一种基于连续时间锁存器的伊辛机(CTLE-Ising),具有1440个自旋,并实现了单次全并行自旋更新,用于高效解决组合优化问题。相比超导量子比特伊辛机需要极低温和高功耗,该设计可在室温下低功耗运行。

💡 主要创新点

重要性
发表年份
ISSCC 2023

🏷 关键词

伊辛机连续时间锁存器全并行自旋更新组合优化

📄 原文摘要

*Equally Credited Authors (ECAs) The Ising machine is a hardware accelerator that finds solutions to combinatorial optimization problems (COPs) using the natural convergence behavior of the Ising model, which comprises artificial-spin network. A previous Ising machine, which is based on the quantum interactions between superconducting quantum bits (qubits) [1], implemented 2048 qubits, but it requires an extremely-low operating temperature (1520mK) and has significant power consumption (10-100kW). Recently, low-power Ising machines [2-4] operating at room temperatures have been developed using low-cost CMOS technologies. They implement discrete-time Ising machines using digital [2,3] or mixed-signal arithmetic circuits [4] to compute the spin operations sequentially. Hence, the performance of these Ising machines is limited by the number of spin-updates, which exponentially increase with the problem size (i.e., the number of binary COP variables).

👥 作者与机构

Jooyoung Bae*, Wonsik Oh*, Jahyun Koo, Bongjin Kim

University of California, Santa Barbara, CA

分类:AI / ML · 年份:ISSCC 2023