Workshop on Addressing Complexity in Multiscale Modeling and Analysis of Complex Data

现代科学的所有分支都需要数据，在过去的三十年里，生物医学研究中的现代数据采集技术已经适应了数据量至少增加了三个数量级的情况。这些数据中的大部分都相当复杂。复杂数据具有三个特性：大容量(例如，基因组数据，每人千兆字节)、多模式(例如，时间x基因x控制)、异构性(例如，不同的时间尺度、数据馈送)。如何分析这些数据现在是科学发现中的一个重大挑战。虽然一系列强大的统计学习算法可以为模式识别问题提供准确的预测，但这些学习算法的能力的增加是以在高维空间中调整大量参数的复杂性为代价的。然而，这种在计算上稳定、在统计上有意义的数学上合理的模型仍然需要比大多数生物实验室和诊所提供的更多的数据。这一研讨会的重点与NSF和NIH之间的倡议相吻合，该倡议旨在研究将基于计算的物理模型与生物学和医学相结合的发展科学和技术。在RPI举行的这个为期两天的研讨会符合NSF和NIH之间的倡议，该倡议旨在研究将基于计算的物理模型与生物学和医学相结合的发展科学和技术。每天结束时都会有一个小组讨论特定的主题，例如如何整合尺度、如何处理不确定性和缺失数据、如何在多尺度建模中使用数据语义。小组讨论将被组织成一份报告，并将向公众公布。研讨会和小组讨论将被记录下来，并向公众公布。

All branches of modern science demand data, and modern data acquisition technologies in biomedical research have adapted to increase the volume of data by at least three orders of magnitude in the past three decades. Much of this data is rather complex. Complex data has three properties: high volume (e.g., genome data, gigabyte per person), multimodal (e.g., Time x Genes x Control), heterogeneous (e.g., different time scales, data feeds).How these data are analyzed now presents a major challenge in scientific discovery. Although a series of powerful algorithms for statistical learning can yield accurate predictions for pattern recognition problems, the increased power of these learning algorithms has come at the expense of a considerable increase in complexity for adjusting a large number of parameters in a high dimensional space. Yet, such mathematically sound models that are computationally stable and statistically meaningful still demand more data than most biology laboratories and clinics can provide. This workshop focus fits into the initiative between NSF and NIH to investigate developing science and technology for merging computational, physics based models with biology and medicine. This two day workshop at RPI fits into the initiative between NSF and NIH to investigate developing science and technology for merging computational, physics based models with biology and medicine. There will be a panel at the end of each day on a particular topic, e.g., how to integrate scales, how to deal with uncertainty and missing data, how to use data semantics in multi-scale modeling. The panel discussions will be organized into a report and will be made publicly available. The seminars and panel discussions will be recorded and made it available to public.