Large Scale Structure Studies on a Value Added Galaxy Catalog

新的统计和机器学习技术将创造一个更好的星系目录。这份目录将结合来自几个现有调查的信息。一项名为“光度红移”的技术将从对星系颜色的研究中获得速度和距离信息。结果将是迄今为止最广泛和最深的目录。它将是比较其他大面积调查，特别是宇宙微波背景（CMB）的理想选择。这些研究将揭示宇宙学和暗能量。它们甚至可能表明是否需要一个更奇特的引力理论。教育包括在专题课程中。与国家实验室有联系。这些方法将介绍给其他学科的学生。该项目将创建一个名为PS1*的组合增值星系目录，用于大规模结构（LSS）的统计研究。基于使用支持向量机（SVM）和来自斯隆数字巡天（SDSS）数据的训练集为第一颗PanSTARRS星表（PS1）开发恒星-星系分离算法的经验，该工作将通过匹配广域红外巡天探测器（WISE）数据、PS1以及SDSS和两微米巡天（2MASS）对象来进行。支持向量机算法将扩展到输出目录。由此产生的星系地图将比SDSS大几倍，比PS1更少的恒星污染，并且将比WISE或2MASS更深。之前演示的机器学习工具将用于估计组合样品的光度红移。新的PS1*将是面积最广、光度最深的红移目录，将是与其他广泛数据集（如CMB、x射线调查、宇宙红外背景和引力透镜图）相互关联研究的最佳选择。将要研究的具体问题包括：CMB异常是否由LSS引起，PS1*中发现的上层结构是否对CMB有任何影响，从PS1*中创建的综合萨克斯-沃尔夫（ISW）效应图如何与普朗克卫星的CMB图相关联，以及LSS统计数据和宇宙学参数可以提取得多好。该项目将探索用于恒星星系分离和光度红移的新型机器学习技术，并应用最先进的统计技术，揭示宇宙参数和暗能量。它将调查LSS和CMB异常之间的任何联系，以及这是否与ISW一致，还是需要一个更奇特的理论。该目录将公开供各种相互关联研究使用，所开发的算法和开放源码软件将广泛传播。本研究透过课程、实习及接纳跨学科学生，促进研究与教学的整合。

New statistical and machine learning techniques will create a better catalog of galaxies. This catalog will combine information from several existing surveys. A technique called "photometric redshifts" will give speed and distance information from study of the galaxy colors. The result will be the widest area and deepest catalog yet available. It will be ideal for comparing to other large area surveys, especially of the Cosmic Microwave Background (CMB). Those studies will shed light on cosmology and dark energy. They may even show whether a more exotic theory of gravity is needed. Education is included through special topic courses. There are connections to national laboratories. The methods will be introduced to students from other disciplines.This project will create a combined value-added galaxy catalog to be called PS1*, for statistical studies of large-scale structure (LSS). Building on experience developing a star-galaxy separation algorithm for the first PanSTARRS catalog (PS1) using Support Vector Machines (SVM) and training sets from Sloan Digital Sky Survey (SDSS) data, the work will proceed by matching Wide-field Infrared Survey Explorer (WISE) data, PS1, and where available, SDSS and Two Micron All Sky Survey (2MASS) objects. The SVM algorithm will extend over the output catalog. The resulting galaxy maps will be several times larger than SDSS, with less stellar contamination than PS1 alone, and will be deeper than WISE or 2MASS. Previously demonstrated machine learning tools will then be used to estimate photometric redshifts for the combined sample. The new PS1* will be the widest area, deepest photometric redshift catalog, and will be optimal for cross correlation studies with other wide data sets, such as the CMB, X-ray surveys, the cosmic infrared background, and maps of gravitational lensing.Specific questions that will be studied include whether CMB anomalies are caused by LSS, whether superstructures that will be found in PS1* have any effect on the CMB, how the map of the Integrated Sachs-Wolfe (ISW) effect to be created from PS1* relates to the CMB map from the Planck satellite, and how well LSS statistics and cosmological parameters can be extracted. The project will explore novel machine learning techniques for both star-galaxy separation and photometric redshifts and apply state of the art statistical techniques, shedding light on cosmological parameters and dark energy. It will investigate any connection between LSS and CMB anomalies, and whether this is consistent with ISW or requires a more exotic theory. The catalog will be publicly available for a variety of cross-correlation studies, and the algorithms and open source software developed will be disseminated widely. This study promotes the integration of research into teaching through courses, internships, and including cross-disciplinary students.