Machine Learning in Science, Technology, Engineering and Mathematics
-
Edited by:
Vigor Yang
The proposed MLSTEM series will include foundational, theoretical, and applications-oriented books, covering fundamentals, practical applications, and computational methodologies. The series will address several key problems:
Fragmented Information: Many existing resources on machine learning (ML) are fragmented, requiring readers to consult multiple disparate sources. Each volume of the M-TEC series offers cohesive, self-contained knowledge, providing readers with a comprehensive and consistent learning experience. This eliminates the hassle of piecing together information from various references.
Integration of Theory and Practice: Current literature often focuses either on the theoretical aspects or the practical applications of ML, but rarely both. The M-TEC series bridges this gap by integrating theory with practice. It helps readers understand how theoretical concepts can be effectively applied in real-world settings, enhancing their ability to implement ML solutions in various domains.
Accessibility: Designed to be accessible to a wide audience, including students, researchers, and professionals, the M-TEC series presents complex ML topics in a clear and concise manner. This approach ensures that readers, regardless of their background, can easily grasp and apply ML concepts. By demystifying complex theories and providing practical examples, the series makes ML accessible and usable for a broader audience.
Overall, the M-TEC series provides a unified, comprehensive resource that simplifies the learning and application of ML, making it an invaluable tool for anyone looking to master this transformative technology.
Author / Editor information
Vigor Yang is a Professor of Aerospace Engineering and a faculty member of the Machine Learning PhD program at the Georgia Institute of Technology. He is also the founding director of the Jame C. Wu Laboratory of Artificial Intelligence for Technology, Engineering, and Computing (ArTEC). His research is at the interface between engineering and data sciences. He is a member of the U.S. National Academy of Engineering, an academician of Academia Sinica, and a foreign member of the Chinese Academy of Engineering and Indian National Academy of Engineering.
Yingjie Liu is a professor in School of Mathematics and a faculty member of the Machine Learning PhD program at the Georgia Institute of Technology. His research is on the development and analysis of numerical methods for solving partial differential equations. The work includes the back and forth error compensation and correction (BFECC) method, central schemes and central discontinuous Galerkin methods on overlapping cells, hierarchical reconstruction (HR) limiting method, and neural networks with local converging inputs (NNLCI).
Xingjian Wang is an associate professor in the Department of Energy and Power at the Tsinghua University. His research areas encompass the interdisciplinary study of engineering science and machine learning, reduced-order modeling, and theories and analyses of complex fluid flows and combustion. He received the Statistics in Physical Engineering Sciences Award from the American Statistical Association in 2019.
While artificial intelligence has made significant strides in imaging and natural language processing, its utilization in engineering science remains relatively new. This book aims to introduce machine learning techniques to facilitate the emulation of complex fluid flows. The work focuses on projection-based reduced-order models (ROMs) that condense high-dimensional data into a low-dimensional subspace by leveraging principal components. Techniques like proper orthogonal decomposition (POD) and convolutional autoencoder (CAE) are utilized to configure this subspace, establishing a functional mapping between input parameters and solution fields. The applicability of POD-based ROMs for spatial and spatiotemporal problems are explored across various engineering scenarios, including flow past a cylinder, supercritical turbulent flows, and hydrogen-blended combustion. To capture intricate dynamics, common POD, kernel-smoothed POD, and common kernel-smoothed POD methods are developed in sequence. Additionally, the effectiveness of POD and CAE in capturing nonlinear features are compared. This book is designed to benefit graduate students and researchers interested in the intersection of data and engineering sciences.
