EAGER: Statistical Learning in Chip

EAGER：芯片中的统计学习

• 批准号: 1247093
• 负责人: Ronald Blanton
• 金额: $ 30万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2012
• 资助国家: 美国
• 起止时间: 2012-09-01 至 2015-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1247093&HistoricalAwards=false
• 关键词: EAGER; Statistical; Learning; Chip

Integrated electronic systems are pervasive in all aspects of our lives, used in everything from computers to mobile phones to automobiles. However, designing and manufacturing systems that both work and are reliable is becoming extremely difficult due to the significant complexity inherent in the underlying technology. In this NSF EAGER project, we will demonstrate how statistical learning in chip (SLIC) can cope with the non-idealities that arise due to imprecise design and fabrication, and the uncertainty that stems from the system's user and operating environment. Specifically, SLIC will enable an integrated system to "learn" optimal operating points across various applications so as to maximize performance, and to minimize power consumption. This will be accomplished by developing customized statistical learning algorithms for in-chip implementation that are capable of deriving actionable information from system data produced both on- and off-line. The principal investigator (PI) is committed to having a broader impact through training a diverse group of undergraduate and graduate researchers. His research group has members from under-represented groups that include women, African Americans, Hispanic Americans, and Native Americans. In addition, as director of the Center for the Silicon System Implementation (CSSI) at Carnegie Mellon University, the PI manages a program that recruits undergrads researchers from various universities (including minority-serving institutions), and the annual convention of the National Society of Black Engineers. This program has been very successful, resulting in the recruitment of many undergraduate researchers, including both women and African-Americans. In the last few years, the PI has supervised nine undergraduate researchers, three of which were African-American (two male and one female), and will continue to recruit a diverse group of students, both at the graduate and undergraduate levels, for participation in this project.

集成的电子系统在我们生活的各个方面都普遍存在，从计算机到手机再到汽车的各个方面都使用。但是，由于基础技术固有的重大复杂性，设计和制造系统都起作用且可靠。在这个NSF急切的项目中，我们将演示芯片中的统计学习如何应对由于不精确的设计和制造而引起的非理想性，以及源于系统的用户和操作环境所带来的不确定性。 具体而言，SLIC将使集成系统能够“学习”各种应用程序的最佳操作点，以最大程度地提高性能并最大程度地减少功耗。这将通过开发用于芯片内实施的自定义统计学习算法来实现，这些算法能够从在线和离线产生的系统数据中获取可行的信息。 首席研究员（PI）致力于通过培训多样化的本科和研究生研究人员群体来产生更大的影响。他的研究小组的成员来自代表性不足的小组，包括妇女，非裔美国人，西班牙裔美国人和美洲原住民。 此外，作为卡内基·梅隆大学（Carnegie Mellon University）的硅系统实施中心（CSSI）主任，PI管理了一项计划，该计划招募了来自各种大学（包括少数派服务机构）的本科生研究人员和黑人工程师的年度大会。该计划非常成功，导致招募了包括女性和非裔美国人在内的许多本科研究人员。在过去的几年中，PI监督了九名本科研究人员，其中三名是非裔美国人（两名男性和一名女性），并将继续招募一群在研究生和本科生的学生参与该项目。