FET: Small: Increasing Robustness, Efficacy, and Capability of CMOS-Compatible Electronic Ising Machines

FET：小型：提高 CMOS 兼容电子发射机的鲁棒性、效率和能力

• 批准号: 2233378
• 负责人: Michael Huang
• 金额: $ 60万
• 依托单位: University of Rochester
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-02-01 至 2026-01-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2233378&HistoricalAwards=false
• 关键词: FET; Small; Increasing; Robustness; Efficacy

Optimization problems are prevalent in modern societies. Solving them quickly and efficiently is a key enabler for continued economic growth and sustainability. Physics-inspired algorithms have long helped scientists and engineers solve these problems. More recently, researchers have gone one step further and used physical processes to directly perform “physical computation”. When researchers can finally bring such a technology to the marketplace, it is anticipated that for some optimization problems, specialized hardware can improve the speed and energy efficiency by thousands or even millions of times. The project will allow the researchers to provide excellent opportunities for both graduate and under-graduate students to participate in cutting-edge inter-disciplinary research and become better prepared to contribute to the future workforce.Based on the foundation of prior work leading to this project, the goal of the team is to address a number of fundamental challenges so as to bring to fruition a physical computing system that can robustly accelerate a broad class of real-world problems. First, the funding will enable the team to build physical prototypes to better understand hardware non-idealities so as to develop cross-layer countermeasures for robust operation. Second, the project will make the first effort in addressing capacity limits and semantic gaps that are two major system-level issues confronting Ising machines. Third, the project will expand Ising machines beyond canonical problems, which can potentially transform them from a niche solution to a mainstream platform in critical application areas such as machine learning.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

优化问题在现代社会中普遍存在。快速有效地解决这些问题是经济持续增长和可持续性的关键推动因素。长期以来，受物理启发的算法一直在帮助科学家和工程师解决这些问题。最近，研究人员更进一步，使用物理过程直接执行“物理计算”。当研究人员最终能够将这种技术推向市场时，预计对于一些优化问题，专用硬件可以将速度和能效提高数千甚至数百万倍。该项目将使研究人员能够为研究生和本科生提供参与前沿跨学科研究的绝佳机会，并为未来的劳动力做好更好的准备。在此项目前期工作的基础上，该团队的目标是解决一些基本的挑战，以便实现一个物理计算系统，加速了一系列现实问题。首先，这笔资金将使团队能够构建物理原型，以更好地了解硬件的非理想性，从而为稳健的操作开发跨层对策。其次，该项目将首次努力解决Ising机器面临的两个主要系统级问题：容量限制和语义差距。第三，该项目将使Ising机器超越规范问题，从而有可能将其从利基解决方案转变为机器学习等关键应用领域的主流平台。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Supporting Energy-Based Learning with an Ising Machine Substrate: A Case Study on RBM

• DOI: 10.1145/3613424.3614315
• 发表时间: 2023-04
• 期刊: 2023 56th IEEE/ACM International Symposium on Microarchitecture (MICRO)
• 作者: Uday Kumar Reddy Vengalam;Yongchao Liu;Tong Geng;Hui Wu;Michael Huang

Leibniz International Proceedings in Informatics (LIPIcs):26th International Conference on Theory and Applications of Satisfiability Testing (SAT 2023)

莱布尼茨国际信息学学报 (LIPIcs)：第 26 届可满足性测试理论与应用国际会议 (SAT 2023)

• DOI: 10.4230/lipics.sat.2023.11
• 发表时间: 2023
• 期刊: Schloss Dagstuhl – Leibniz-Zentrum für Informatik
• 作者: Haberlandt, Andrew;Green, Harrison;Heule, Marijn J.
• 通讯作者: Heule, Marijn J.