CAREER: Resource Efficient Systems for Machine Learning on Structured Data

职业：结构化数据机器学习的资源高效系统

• 批准号: 2237306
• 负责人: Shivaram Venkataraman
• 金额: $ 67.61万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-07-01 至 2028-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2237306&HistoricalAwards=false
• 关键词: CAREER; Resource; Efficient; Systems; Machine

Many scientific and enterprise datasets include relationships among data items, which can be represented as graphs. Applying machine learning methods on such graph datasets can yield benefits across several domains, including social networks, drug discovery, and search engines. However, existing software for applying machine learning methods on large graph datasets is slow, complex, and expensive. This NSF CAREER proposal aims to address these challenges by developing software that will make it faster, easier, and less expensive to analyze large graph datasets. The proposed research includes three thrusts that focus on different stages of machine learning workflows. The first thrust aims to develop software that will make it faster and easier to train machine learning models on large graphs using many machines. The second thrust focuses on how to efficiently handle scenarios where graphs are updated with additional data. The third thrust considers how to make prediction less expensive when using machine learning models trained on large graph datasets. The broader impacts of the proposed research include improved analysis capabilities for data scientists working in many areas. Furthermore, all software developed as a part of this project will be made freely available to the wider community and will include documentation and tutorials to help users from computer science and other academic disciplines to get started. Additionally, the proposal plans to develop a new undergraduate course that teaches students how to use software frameworks to process large datasets. The assignments in the course will use software tools developed as a part of this project. The project also includes plans to broaden participation in computer science by organizing a yearly workshop that promotes research opportunities for undergraduate students from underrepresented groups, as well as discussion sessions that can help students who are getting started with research.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.

许多科学和企业数据集包括数据项之间的关系，这些关系可以表示为图形。在这些图形数据集上应用机器学习方法可以在多个领域产生效益，包括社交网络、药物发现和搜索引擎。然而，现有的在大型图数据集上应用机器学习方法的软件是缓慢、复杂和昂贵的。NSF CAREER提案旨在通过开发软件来解决这些挑战，这些软件将使分析大型图形数据集更快，更容易，更便宜。拟议的研究包括三个重点，重点关注机器学习工作流程的不同阶段。第一个目标是开发软件，使其更快，更容易地使用许多机器在大型图上训练机器学习模型。第二个重点是如何有效地处理图表使用额外数据更新的场景。第三个重点是考虑如何在使用在大型图数据集上训练的机器学习模型时降低预测成本。拟议研究的更广泛影响包括提高在许多领域工作的数据科学家的分析能力。此外，作为该项目一部分开发的所有软件将免费提供给更广泛的社区，并将包括文档和教程，以帮助计算机科学和其他学科的用户入门。此外，该提案计划开发一门新的本科课程，教授学生如何使用软件框架来处理大型数据集。课程中的作业将使用作为本项目一部分开发的软件工具。该项目还包括通过组织年度研讨会来扩大计算机科学参与度的计划，该研讨会为来自代表性不足群体的本科生提供研究机会，以及可以帮助开始研究的学生的讨论会。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。