CISE Research Instrumentation: Equipment in Support of Parallel Processing Research

9729858富兰克林，马克A. 华盛顿大学 CISE研究仪器：支持并行处理研究的设备 这种研究仪器可以在以下领域进行研究：-模式和图像识别，-计算生物学和-离散事件模拟。 仪器补助金为购买Utraplant II处理器、数据存储器、接口、备份磁盘和软件提供资金。并行处理器将支持需要大量计算资源的研究，以维持当前的努力并取得更多进展。下文简要介绍了这些项目。 - 乳房X光检查：将继续研究开发从乳房X光检查中发现癌症的自动化方法。我们的目标是改善现有的并行实现有效的临床使用。所采用的技术结合了人工神经网络和小波变换技术的联合收割机。 - MCD（多重完全消化）图谱：将继续研究一种高分辨率的限制性片段DNA图谱（MCD图谱）。作图问题是基于DNA的遗传学研究的基础。 由于基因组的规模巨大（数百万到数十亿个碱基对），即使是相对简单的支持活动，如软件测试，也非常耗时。所要求的设备将用于开发更快和更准确的并行算法。 - 离散事件模拟：将继续研究开发并行离散事件模拟算法，用于模拟大型计算机通信网络。 有效 需要网络设计来持续扩展和改进因特网和万维网。并行模拟技术将用于评估系统性能时，相对罕见的事件发生。 例如，中等规模网络中的阻塞概率评估需要处理超过10万亿个事件。

9729858 Franklin, Mark A. Washington University CISE Research Instrumentation: Equipment in Support of Parallel Processing Research This research instrumentation enables research in the following areas:- Patterns and Image Recognition,- Computational Biology, and- Discrete Event Simulation. The instrumentation grant funds the purchase of UtraSPARC II processors, data storage, interfaces, backup disk, and software. The parallel processor will support research requiring considerable computational resources to sustain current efforts and make additional progress. A brief description of the projects follows. - Mammogram Screening: Research will continue in the development of automated means for detecting cancer from mammograms. The goal is to improve existing parallel implementations for effective clinical usage. The techniques employed combine the technologies of artificial neural networks and wavelet transforms. - MCD (Multiple Complete Digestion) Mapping: Research will continue on a form of high-resolution, restriction-fragment DNA mapping (MCD mapping). The mapping problem is fundamental to DNA based genetic research. With genomes of enormous size (millions to billions of base pairs), even relatively simple support activities such as software testing are extremely time consuming. The requested equipment will be used to develop parallel algorithms which are faster and more accurate. - Discrete-Event Simulation: Research will continue on developing parallel discrete-event simulation algorithms for modeling large computer communications networks. Effective network designs are needed for the continued expansion and improvement of the Internet and World Wide Web. Parallel simulation techniques will be used to evaluate system performance when relatively rare events occur. For example, blocking probability evaluation in modest sized networks requires processing more than 10 trillion events.