MRI: Acquisition of Heterogeneous Computer System for Machine Learning

MRI：获取用于机器学习的异构计算机系统

• 批准号: 1919752
• 负责人: Steven Skiena
• 金额: $ 57.2万
• 依托单位: SUNY at Stony Brook
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1919752&HistoricalAwards=false
• 关键词: MRI; Acquisition; Heterogeneous; Computer; System

This award supports the purchase, deployment, and operation of a heterogeneous computer system to advance the research interests of many faculty in the Stony Brook University's Artificial Intelligence research community. The shared research instrument will enable discovery in a multitude of areas. For example: in Geosciences, the cluster will be used to process satellite imagery to map remote parts of the Arctic and Antarctic; in Biosciences it will be employed to build neural networks that will advance cardiac imagery and fMRI research; in Social Sciences it will help produce methods and tools that can illuminate social and psychological determinants of population health. The provisioning of the instrument in the Stony Brook University campus will greatly benefit hundreds of graduate and undergraduate students taking courses in machine learning and artificial intelligence each year. The proposed cluster will offer professionally-managed resources providing a standardized environment with adequate technical support. Further, student demand is anticipated to continue to grow through several ongoing initiatives including an undergraduate CS concentration in artificial intelligence, with new courses in NLP and data science, a recently approved graduate CS concentration in Data Science, and a proposed "Data Science+X" undergraduate sequence integrating with several majors in the College of Arts and Sciences (CAS).The cluster will comprise of three categories of nodes: (1) a large Graphics Processing Unit (GPU) machine, (2) GPU server machines with large amounts of memory and (3) Field Programmable Gate Arrays (FPGA) machines thus providing a forward-looking computing environment focused on supporting research on and enabled by AI. It will constitute the research instrumentation catalyst for the Institute for AI-Driven Discovery and Innovation (AI Institute) and significantly overhaul the compute capability of the institution. This system?s heterogeneity will allow for a more versatile shared environment with excellent performance. will support research in three directions: 1) Artificial Intelligence (AI) including Natural Language Processing, Computer Vision, and Machine Learning (ML) 2) applications of AI and ML on a variety of scientific disciplines including Biomedical Informatics, Chemistry, Ecology, and Linguistics, 3) Computer Science (CS) research on HPC such as visualization, neural network compilation, and process scheduling. Furthermore, the strong educational and outreach activities that engage high school students and teachers, as well as undergraduate and graduate students have the potential to support the building of a skilled workforce of developers, researchers, staff and users.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.

该奖项支持购买、部署和操作不同种类的计算机系统，以促进石溪大学人工智能研究社区许多教师的研究兴趣。共享的研究仪器将使人们能够在许多领域进行发现。例如：在地球科学方面，该群组将被用于处理卫星图像，以绘制北极和南极偏远地区的地图；在生物科学方面，它将被用于建立神经网络，从而推动心脏成像和功能磁共振研究；在社会科学方面，它将帮助产生能够阐明人口健康的社会和心理决定因素的方法和工具。在石溪大学校园提供该仪器，将使每年数百名正在学习机器学习和人工智能课程的研究生和本科生大大受益。拟议的专题组将提供由专业人员管理的资源，为标准化环境提供充分的技术支持。此外，通过几项正在进行的计划，学生的需求预计将继续增长，包括本科生集中在人工智能，新课程在NLP和数据科学，最近批准的研究生CS集中在数据科学，以及拟议的“数据科学+X”本科生序列整合在艺术和科学学院(CAS)的几个专业。集群将由三类节点组成：(1)大型图形处理单元(GPU)机器，(2)具有大量内存的GPU服务器机器和(3)现场可编程门阵列(FPGA)机器，从而提供一个专注于支持人工智能研究和由其启用的前瞻性计算环境。它将构成人工智能驱动的发现和创新研究所(AI Institute)的研究仪器催化剂，并大幅改革该机构的计算能力。这个系统-S的异构性将允许一个更多功能的共享环境和出色的性能。将支持三个方向的研究：1)人工智能(AI)，包括自然语言处理、计算机视觉和机器学习(ML)；2)AI和ML在各种科学学科中的应用，包括生物医学信息学、化学、生态学和语言学；3)计算机科学(CS)对高性能计算的研究，如可视化、神经网络编译和进程调度。此外，吸引高中生和教师以及本科生和研究生的强有力的教育和推广活动有可能支持开发人员、研究人员、工作人员和用户的熟练劳动力队伍的建设。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势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

Robust Task Representations for Offline Meta-Reinforcement Learning via Contrastive Learning

• DOI: 10.48550/arxiv.2206.10442
• 发表时间: 2022-06
• 期刊: ArXiv
• 作者: Haoqi Yuan;Zongqing Lu

ViTag: Online WiFi Fine Time Measurements Aided Vision-Motion Identity Association in Multi-person Environments

• DOI: 10.1109/secon55815.2022.9918171
• 发表时间: 2022-09
• 期刊: 2022 19th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON)
• 作者: Bryan Bo Cao;Abrar Alali;Hansi Liu;Nicholas Meegan;M. Gruteser;Kristin J. Dana;A. Ashok;Shubham Jain

Subset Node Anomaly Tracking over Large Dynamic Graphs

• DOI: 10.1145/3534678.3539389
• 发表时间: 2022-05
• 期刊: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
• 作者: Xingzhi Guo;Baojian Zhou;S. Skiena

BioNLI: Generating a Biomedical NLI Dataset Using Lexico-semantic Constraints for Adversarial Examples

BioNLI：使用对抗性示例的词汇语义约束生成生物医学 NLI 数据集

• DOI: 10.18653/v1/2022.findings-emnlp.374
• 发表时间: 2022
• 期刊: Findings of the Association for Computational Linguistics: EMNLP 2022
• 作者: Bastan, Mohaddeseh;Surdeanu, Mihai;Balasubramanian, Niranjan
• 通讯作者: Balasubramanian, Niranjan

CORA: Benchmarks, Baselines, and Metrics as a Platform for Continual Reinforcement Learning Agents

• 发表时间: 2021-10
• 期刊: ArXiv
• 作者: Sam Powers;Eliot Xing;Eric Kolve;Roozbeh Mottaghi;A. Gupta