ITR: Searching for Correlations in a High Dimensional Space

AST-0312498康诺利该项目将创建多维数据集的分析工具，以解决当前和计划的天空调查中预期的信息泛滥问题，并开始理解本地和遥远宇宙的全色视图。 利用统计学和新的计算技术的最新进展将克服早期方法的局限性，并导致快速和有效的工具提取相关性。 天体物理数据，而不是模拟，将用于测试这些新的算法，这将涉及依赖树，非线性降维方案和贝叶斯网络。 将开发的方法将在需要分析大量数据集的所有领域产生更广泛的影响。 对研究生进行培训，使他们像熟悉天体物理问题一样熟悉最新的统计或计算机科学技术，将产生教育影响。

AST-0312498ConnollyThis project will create analysis tools for multi-dimensional data sets, in order to address the problem of the expected flood of information from current and planned surveys of the sky, and to start making sense of the resultant panchromatic view of the local and distant Universe. Utilizing recent advances in statistics and new computational techniques will overcome the limitations of earlier approaches and lead to rapid and efficient tools for extracting correlations. Astrophysical data, not simulations, will be used to test these new algorithms, which will involve dependency trees, non-linear dimensionality reduction schemes, and Bayesian networks. The methods to be developed will have a broader impact in all areas where massive data sets need analysis. The training of graduate students to be as comfortable with the latest statistical or computer science techniques as they are with astrophysical problems will have an educational impact.