Research

Simulations are playing an increasingly important role in a wide range of scientific disciplines. While the field of computer science is responsible for formulating many of these simulations, they are utilized pervasively in studies that pertain to healthcare, physics and engineering, to offer just a few examples of the essential areas of research that benefit from advances in simulation models. In the study of computer science, I have proven myself to be an expert in the construction and improvement of simulations through the research endeavors.

I completed work on my research study that used model-order reduction to solve numerical modeling problems that result from an overabundance of information in complex datagenerating systems and simulations. I stressed the importance of consistency and time efficiency in constructing such a solution to this problem in dynamical systems. In an effort to preserve the main features of the numerical modeling problems, I combined innovative compression techniques and optimization methods with gappy-data approximation to form a cost-effective computational approach. The novel model-order reduction methodology that I constructed represented a more time-efficient, cost-efficient system that offered reliable assessments for remediating the cumbersome nature of complex data-generating applications.

I also contributed important work on the topic of uncertainty in solvable computational models. I began my work in this research project by deploying interval methods to address the uncertainty in models while assessing the efficiency of the methods throughout the process of solving the modified problem. My ensuing experiments with logistics equations and simulations with multiple data points of uncertain information allowed my to quantify the association shared by the quantity of the data and the quality of the data, along with the quality of the prediction. Through this process, I noted that the uncertainty with initial time had a correlative relationship to uncertainty in the initial condition. I also used randomized data selection where uncertainty was found to discover that the interval of the initial condition was calculable and assisted in advancing the calculation of advanced nonlinear high-performing computing algorithms.

In my postdoctoral position in the University of Tennessee in Knoxville, I implemented SOMOSPIE to predict soil moisture in a finer resolution. SOMOSPIE takes as input the soil moisture over large areas from remote sensing, such as satellites with radar sensors, providing nearly global coverage. In addition, SOMOSPIE uses meaningful terrain hydrology parameters, which are extracted from Digital Elevation Model (DEM). SOMOSPIE (SOil MOisture SPatial Inference Engine) developed in the Global Computing Lab consists of modular stages for processing spatial environmental data, generating predictions with machine-learning techniques, and analyzing these predictions. The soil moisture of any large region of interest can be estimated using SOMOSPIE. The modeling functionalities of SOMOSPIE include options for data preprocessing and prediction method selection. Data preprocessing capabilities include training domain selection and data dimension reduction. For modeling method selection, our inference engine includes standard machine learning methods based on kernels (i.e., KKNN) and regression trees (i.e., Random Forests or RF).

Finally, based on the importance of my work, the National Science Foundation funded my research with grants. I also received funding through the Army High-Performance Computing Research Center at Stanford University.