IBM SP2 COMPUTER FOR EXPERIMENTAL NEUROCOMPUTATION

所有NIH资助的神经科学家在这个提案中生成，建模， 分析来自复杂神经系统的数据 他们跨学科工作 线和每一个已经开始联合收割机实验和计算相结合。 有 有两个潜在的原因。 第一，理解 最近，他们的研究问题的适当解决方案， 概念上的突破 第二个原因是， 技术上可以执行广泛的和更困难的 测试这些解决方案所需的计算。计算中的问题 神经科学通常涉及许多偶然相互作用的部分。为 例如，在皮层模拟中，神经元的模型可以包括： 数以千计的隔间。大脑某个区域的模拟可以 涉及数百万神经元之间的多重偶然相互作用。 这种组合爆炸导致了巨大的计算需求; 过去30年的计算机器未能满足的需求 几十年 这项建议的研究重点是使用 来批判性地测试神经科学问题。这笔赠款 主要要求资金用于进一步发展高级 性能，神经科学特定的并行计算组件。 请求提供资金，用于购买 IBM SP2并行计算机、额外的磁盘存储和磁带存档。 还要求提供资金，以提供特定模式的高速自动柜员机 网络访问实验室工作台。 高性能并行 计算机将允许教师，博士后学员和毕业生 让学生们计算合理大小的神经科学问题， 结合实验程序。 重要的是， 本地计算能力将允许在线模型和数据操纵， 使用可视化引擎并行计算 发动机高速网络将允许集成 实验数据的并行计算和可视化 收藏.

All the NIH funded neuroscientist in this proposal generate, model, and analyze data from complex neural systems. They work across disclipinary lines and each has begun to combine experiment and computation. There are two underlying reasons for this. First, the understanding of the appropriate solutions to their research questions has recently had conceptual breakthroughs. The second reason is that it is now technically possible to perform the extensive and more difficult computations necessary to test these solutions. Problems in computational neuroscience generally involve many contingently interactive parts. For example in cortical simulations, a model of the neuron can involve thousands of compartments. The simulation of a region of the brain can involve multiple contingent interactions between millions of neurons. This combinatorial explosion leads to tremendous computational demands; demands that were not met by computational machines of the last threc decades. The research studies of this proposal focus on the use of computations to critically test neuroscience questions. This grant primarily requests funds for the further development of the high performance, Neuroscience specific parallel computational components. Funds are requested for purchasing additional computational nodes on the IBM SP2 parallel computer, additional disk storage, and tape archiving. Funds are also requested to provide paradigm-specific high speed ATM network access to the laboratory bench . The high-performance parallel computing will permit the faculty, post-doctoral trainees, and graduate students to compute reasonably sized neuroscience problems, and do this in conjunction with experimental procedures. Importantly the significant local computational power will permit on-line model and data-steering via the use of a visualization engine in parallel with the computational engine. The high Speed network will permit the integration of the parallel computation and visualization with the experimental data collection.