CISE Research Resources: Information Visualization and Incremental Knowledge Discovery in a Cluster Computing Environment

EIA-0130857 Elise deDoncker西密歇根大学CISE研究资源：集群计算环境中的信息可视化和增量知识发现该提案寻求资金，以增强西密歇根大学的64节点PC集群，部分获得了以前的国家科学基金会拨款和大学资金。对集群的改进将为并行分布式算法设计和开发的研究和教育提供硬件支持。 虽然集群目前用于各种研究工作，由不同的团体在大学，本建议侧重于其中两个项目：并行集成和分布式增量知识发现。所要求的基础设施包括网络升级，以支持目前通信密集型（高带宽/低延迟）项目。 这些包括自适应分区中区域迁移的负载平衡方法的定制。 该系统也将为我们在数值积分等领域研究任务划分方法的可扩展性提供更好的环境。 拟议的关于制定从大型数据库中逐步发现知识的方法的项目将进一步得到基础设施的可视化部分的支持。 这项工作包括大型关系数据集的3D交互式可视化探索。 这些技术将以系统的方式纳入一个软件系统，以进行增量知识发现，并评估其在不同情况下的有效性和适用性。

EIA-0130857 Elise deDonckerWestern Michigan UniversityCISE Research Resources: Information Visualization and Incremental Knowledge Discovery in a Cluster Computing EnvironmentThis proposal seeks funding to enhance a 64-node PC cluster at Western Michigan University, obtained in part on a previous National Science Foundation grant and university funding. Improving the cluster will give hardware support for on-going research and education on parallel and distributed algorithm design and development. Although the cluster is currently used for various research efforts by different groups in the university, this proposal focuses on two of these projects: parallel integration and distributed incremental knowledge discovery. The requested infrastructure includes a networking upgrade to support the current communication intensive (high bandwidth/low latency) projects. These include the tailoring of load balancing methods underlying the migration of regions in adaptive partitioning. The system will also provide a better environment for our study of the scalability of task partitioning methods in numerical integration and other areas. The proposed project on the development of methods for incremental knowledge discovery from large databases will further be supported by the visualization components of the infrastructure. This effort includes 3D interactive visual exploration of large relational data sets. The techniques will be incorporated in a systematic way into a software system for incremental knowledge discovery, and evaluated for effectiveness and suitability in different scenarios.