II-New: Infrastructure for Supporting Biomedical Application Algorithms, Runtime Development and Resource Management

由基金资助的计算基础设施支持的研究侧重于生物医学应用，这些应用必须(I)处理巨大尺寸的图像或图形数据，(Ii)处理具有噪声和变形的数据，以及(Iii)需要巨大的计算能力。正在研究的一个例子是使用大型3D显微荧光图像研究乳腺癌脑转移。这样的生物医学应用可以从适当的异质计算基础设施和从算法开发到算法到计算基础设施的有效映射的协调努力中受益匪浅。此外，该基础设施将为实施足够严谨的CS培训计划提供平台，供教师、博士后研究人员和学生与同行机构共享，并将被整合到多个现有的面向代表不足的少数族裔、高中教师和学生的外展计划中，如圣母大学夏季学者研究计算轨道和NSF REU计算站点计划。这项工作的成功完成将推进解决癌症和其他疾病治疗中的基本问题的计算方法，并发现模仿生物系统感觉处理和推理能力的新算法类别。这项工作还将提供强大的可视化和图形分析技术，用于处理大型和复杂的图像/图形数据并进行推断，这对许多生物医学和其他应用程序至关重要。最后，它将导致应用程序驱动的资源管理机制的发展，这些机制对于有效利用任何基于GPU的异构计算基础设施是不可或缺的。

The research supported by the funded computing infrastructure focuses on biomedical applications that must (i) process images or graph data of enormous size, (ii) handle data with noise and deformation, and (iii) require tremendous computational capability. One example problem being investigated is studying breast cancer brain metastasis using large 3D microscopy fluorescence images. Such biomedical applications can benefit greatly from both appropriate heterogeneous computing infrastructure and concerted effort spanning from algorithm development to effective mapping of algorithms to the computing infrastructure. Furthermore, the infrastructure will provide a platform for implementing a CS training plan of sufficient rigor for faculty, postdoctoral researchers and students to share with peer institutions, and will be integrated into multiple existing outreach programs to underrepresented minority, high school teachers and students such as the Notre Dame Summer Scholars Research Computing Track and the NSF REU Site program in Computational Science. A top-down, vertically integrated theme is adopted in the proposed research effort. Successful completion of the work will advance computational approaches for tackling fundamental problems in treating cancer and other diseases, and for discovering new classes of algorithms that mimic the sensory processing and reasoning abilities of biological systems. The effort will also offer powerful visualization and graph analysis techniques for processing and making inferences from large and complex image/graph data, which are essential for many biomedical and other applications. Finally, it will lead to the development of application-driven resource management mechanisms that are indispensable for effectively exploiting any GPU-based heterogeneous computing infrastructure.