Learning equations of motion for coupled spatio-temporal processes

点击翻译按钮获取中文摘要

We will develop and validate methods for learning the equations of motion associated with microstructure growth in alloys. Such processes are described by coupled partial differential equations of motion, each describing the local concentration or degree of crystallinity of one of the components. The equations proposed theoretically have become increasingly complex, making interpretation and validation challenging. By using a data driven approach to learning the equations, we hope to demonstrate that the methods of machine learning can complement more traditional approaches to developing theoretical descriptions of spatio-temporal processes.