CAREER: An Efficient Framework for Design and Modeling of Complex Computer Experiments

本研究的主要目标是为复杂计算机实验的设计和建模开发一个有效的框架，特别是那些具有不同和高维输入和大量输出的实验。计算机实验，即使用模拟或数字编码的实验，已被广泛用作物理实验的替代方法，特别是用于研究复杂现象。研究者介绍了新的设计类，可以有效地适应大量的定量和定性因素在计算机实验。研究者还提出了一种新的适应性设计，该设计灵活而稳健，但考虑到复杂系统中的不确定性。除了实验设计之外，本文还提出了一种新的采样/建模技术，以降低大量数据计算机实验分析的计算复杂度和量化模型不确定性。所提出的方法很容易适用于各种科学学科，并将对加速涉及生物力学工程、系统生物学和环境科学等复杂实验的许多领域的发现产生直接影响。特别是，所提出的建模技术可以显著提高细胞粘附分析的效率和预测精度，而细胞粘附在癌症研究中对肿瘤转移起着重要的作用。此外，所提出的方法有利于分析气候变化、应对自然灾害和流行病传播的大量数据。该提案中概述的研究中包含了一项教育计划，该计划强调为广大学生提供跨学科培训，并增加代表性不足群体的参与。拟议的研究结果将被纳入罗格斯大学提供的本科生研究经验项目。来自代表性不足群体的女性和本科生将被招募，并通过罗格斯大学非常成功的科学与工程研究项目积极参与PI的研究。将编写软件，使研究生和本科生有实践经验，在实际例子中实现新方法。

The primary objective of this research is to develop an efficient framework for design and modeling of complex computer experiments, especially those with diverse and high-dimensional inputs and massive outputs. Computer experiments, i.e., experiments using simulation or numerical codes, have been widely used as alternatives to physical experiments, especially for studying complex phenomena. The investigator introduces new classes of designs that can efficiently accommodate large numbers of both quantitative and qualitative factors in computer experiments. The investigator also proposes a new adaptive design that is flexible and robust, yet takes into account uncertainties in the complex systems. Apart from experimental design, a novel sampling/modeling technique is proposed to reduce computational complexity and quantify model uncertainty in the analysis of computer experiments with massive data.The proposed approaches are readily applicable to a variety of scientific disciplines and will have immediate impact on accelerating discoveries in numerous fields involving complex experiments like biomechanical engineering, systems biology, and environmental science. In particular, the proposed modeling techniques can dramatically enhance the efficiency and prediction accuracy in the analysis of cell adhesion, which plays an important role in tumor metastasis in cancer research. Moreover, the proposed methods can benefit the analysis of massive data for climate change, response to natural disasters, and the spread of pandemic disease. Integrated into the research outlined in this proposal is an education plan that emphasizes interdisciplinary training for a broad body of students and increasing participation from underrepresented groups. Results of the proposed research will be integrated into the Research Experience for Undergraduates program offered by Rutgers. Female and undergraduate students from underrepresented groups will be recruited and actively involved in the PI's research through Rutgers's highly successful Research in Science and Engineering program. Software will be written, which allows graduate and undergraduate students to have hands-on experience to implement the new methods on real examples.