KDI: Computational Challenges in Cosmology

在过去的十年里，宇宙学经历了一次复兴，从一门数据匮乏的科学转变为数据驱动的科学。COBE卫星和随后对宇宙微波背景（CMB）的观测已经开始为我们提供早期宇宙的详细图像；望远镜已经发现了距离相当于宇宙十分之一年龄的星系；大规模的红移观测已经开始绘制出附近宇宙的结构。然而，这些数据集的规模可能会使宇宙学数据不堪重负。实现我们的科学目标取决于满足定量新数据带来的定性新计算挑战。宇宙学家在数据的分析、合成和呈现方面面临的问题——包括数据压缩和传输、海量存储、数据挖掘、并行算法和缩放、逆问题和正则化方法，以及复杂数据可视化——也是当前计算机科学和统计学研究的前沿。研究人员和他们的同事在天体物理学家、统计学家和计算机科学家之间形成了一个集中的合作，以开发计算工具、技术和技术，以应对这些数据集带来的新挑战。宇宙学的需求为计算机科学和统计学的新发展提供了实际的刺激，使天体物理学和其他更广泛的数据密集型学科的新研究成为可能并为其提供信息。研究沿着四条环环相扣的路径进行。首先，必须有合适的工具来分析单个数据集。接下来是对数据集的综合和模拟，以形成宇宙结构演化的连贯图景。最后，必须有访问数据和分析结果的权限，协作成员和外部社区都可以访问。宇宙学是对最大尺度宇宙的理解，以及对大爆炸后最初时刻发生的事件的探索。现在构成宇宙微波背景（或CMB）的光最后一次与宇宙中的物质相互作用是在它大约10万年的时候（与它今天150亿年的年龄相比，这只是一小部分）；观测CMB可以让宇宙学家绘制出宇宙早期的地图。在离我们更近的地方，观察星系的分布可以让我们看到宇宙的现状。就在十年前，这些研究中涉及的数据量很小。望远镜和探测器技术的进步使这些数据增加了许多倍；这种巨大的扩张推动了我们分析它的计算能力的极限。来自天体物理学、计算机科学和统计学的研究人员和他们的同事开发了分析和综合这些大量数据的工具。这使他们能够形成一个连贯而完整的宇宙演化图景，从最早的时代到现在，也许最重要的是，到遥远的未来。

Silk9872979 In the past decade cosmology has undergone a renaissance,transforming from a data-starved science to a data-driven one.The COBE satellite and subsequent observations of the CosmicMicrowave Background (CMB) have begun to give us a detailedpicture of the early Universe; telescopes have found galaxies atdistances corresponding to the Universe at one-tenth of itspresent age; large-scale redshift surveys have begun to map outthe structure of the nearby Universe. However, the size of thesedatasets threatens to leave cosmology data-swamped. Realizing ourscientific goals depends on meeting the qualitatively newcomputational challenges set by the quantitatively new data. Theissues cosmologists face in the analysis, synthesis andpresentation of the data --- including data compression andtransmission, mass storage, data mining, parallel algorithms andscaling, inverse problems and regularization methods, and complexdata visualization --- are also at the forefront of currentresearch in Computer Science and Statistics. The investigatorsand their colleagues form a focused collaboration betweenastrophysicists, statisticians and computer scientists to developcomputational tools, techniques and technologies to cope with thenew challenges posed by these datasets. The needs of Cosmologyprovide a practical spur to new developments in Computer Scienceand Statistics, enabling and informing new research both withinastrophysics and more widely in other data-intensive disciplines.Research is organized along four interlocking paths. First, theremust be appropriate tools to analyze the individual datasets.Next are the synthesis and simulation of the datasets toformulate a coherent picture of the evolution of structure in theUniverse. Finally, there must be access to the data and theproducts of the analysis, both to members of the collaborationand to the outside community. Cosmology is the quest for the understanding of the Universeon the largest scales, and of the events that unfolded in thefirst moments after the Big Bang. Light that now makes up theCosmic Microwave Background (or CMB) last interacted with matterin the Universe when it was about one hundred thousand years old(a small fraction of its age today of fifteen billion years);observing the CMB allows cosmologists to map out the Universe atthese very early times. Somewhat closer to home, observing thedistribution of galaxies lets us see the present state of theUniverse. Only ten years ago, the amount of data involved inthese studies was tiny. Advances in telescope and detectortechnology has allowed a manyfold increase in these data; thisvast expansion pushes the limits of our computational ability toanalyze it. The investigators and their colleagues, fromastrophysics, computer science, and statistics, develop tools toanalyze and synthesize this vast amount of data. This allows themto form a coherent and complete picture of the evolution of theUniverse from the earliest times to the present day and ---perhaps most importantly --- far into the future.