CIF: Small: Adaptive Information: Sequential Sensing and Active Learning Theory, Methods and Applications

复杂的系统，如活细胞和互联网，涉及大量单独组件之间的相互作用和关联。具体的例子包括生物细胞中蛋白质之间的相互作用以及通信网络中用户之间的连接。发展大规模网络系统建模和分析的理论和方法是21世纪的重大挑战。例如，新的数学方法将导致更好的生物系统模型，从而改进疾病预防。测量这类系统的信号是开发良好模型的关键一步，而一个主要障碍是，测量大型系统中的所有变量往往是不切实际的，甚至是不可能的。该项目解决了这一障碍，开发了自适应方法，利用从以前收集的数据中收集的信息自动调整测量过程，以便集中和优化新信息的收集。最终，这些方法将极大地加快科学和工程领域的发现步伐。该项目的主要主题是调查自适应测量、传感和实验在具有多个变量的大型复杂系统中的作用。自适应方法是基于先前收集的数据优化下一次测量或实验选择的顺序程序。这项研究涉及发展一种适用于工程和科学各个领域的自适应测量的一般理论。主要目标是从数学上描述和量化自适应测量相对于非自适应方法的优势，并设计最优的自适应测量程序。调查还通过研究生物学、国家安全和人机交互中的具体问题和应用来探索自适应方法的潜力。

Complex systems, such as living cells and the Internet, involve interactions and associations between large numbers of individual components. Specific examples include interactions between proteins in biological cells and connections between users in communication networks. Developing theory and methods to model and analyze large-scale networked systems is a grand challenge for the 21st century. For instance, new mathematical methods will lead to better models of biological systems and thus improvements in disease prevention. Measuring the signals of such systems is a crucial step in developing good models, and a major hurdle is that it is often impractical or impossible to measure all variables in large systems. This project addresses the hurdle by developing adaptive methods that automatically adjust the measurement process by using information gleaned from previously collected data in order to focus and optimize the gathering of new information. Ultimately, these methods will dramatically accelerate the pace of discovery in science and engineering.The main theme of the project is an investigation of the role of adaptive measurement, sensing and experimentation in large complex systems of many variables. Adaptive methods are sequential procedures that optimize the selection of the next measurements or experiments based on previously gathered data. The research involves the development of a general theory for adaptive measurement that is applicable to various domains of engineering and science. The main goals are to mathematically characterize and quantify the advantages of adaptive measurements relative to non-adaptive methods and to design optimal adaptive measurement procedures. The investigation also explores the potential of adaptive methods by studying specific problems and applications in biology, national security, and human-computer interaction.