CSR: Small: Dynamic Construction and Configuration of Classifier Topologies for Real-time Stream Mining Systems

CSR：小型：实时流挖掘系统的分类器拓扑的动态构建和配置

• 批准号: 1016081
• 负责人: Mihaela van der Schaar
• 金额: $ 29.99万
• 依托单位: University of California-Los Angeles
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2010
• 资助国家: 美国
• 起止时间: 2010-09-15 至 2014-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1016081&HistoricalAwards=false
• 关键词: CSR; Small; Dynamic; Construction; Configuration

This project is investigating the following three research problems concerning knowledge extraction and data mining from real-time stream systems:Novel Framework for Stream Mining Systems: The project is studying how to formalize the problem of large scale distributed knowledge extraction from high volume streams by defining appropriate local and end-to-end objective functions, along with resource and delay constraints that will guide the different optimization and adaptation algorithms. Topology construction: The project is studying methods to organize classifiers into a connected topology mapped onto a distributed infrastructure. The research starts by studying linear chains of classifiers and then seeks to extend the work to multiple chains working in parallel, and to classifier trees.Decentralized Solutions based on Interactive Multi-Agent Learning: For large scale stream mining systems, where the classifiers are distributed across multiple nodes, the project is developing a decentralized decision framework and distributed algorithms for joint topology construction and local classifier configuration. In such distributed scenarios, optimizing the end-to-end performance requires interactive, multi-agent solutions to be deployed at each site in order to determine the effect of each classifier's decisions on the other classifiers.

该项目正在研究以下三个有关实时流系统的知识提取和数据挖掘的研究问题：流挖掘系统的新颖框架：该项目正在研究如何通过定义从大容量流中形式化大规模分布式知识提取的问题适当的本地和端到端目标函数，以及指导不同优化和适应算法的资源和延迟约束。拓扑构建：该项目正在研究将分类器组织成映射到分布式基础设施的连接拓扑的方法。该研究从研究分类器的线性链开始，然后寻求将工作扩展到并行工作的多个链以及分类器树。基于交互式多代理学习的去中心化解决方案：对于分类器是分布式的大规模流挖掘系统该项目正在开发一个跨多个节点的去中心化决策框架和分布式算法，用于联合拓扑构建和本地分类器配置。在这种分布式场景中，优化端到端性能需要在每个站点部署交互式多代理解决方案，以确定每个分类器的决策对其他分类器的影响。