Scientific Machine Learning Lab
TAMIDS has created a new program to collaborate with faculty teams in establishing Labs in emerging areas of Data Science. The mission of each Lab will be to develop knowledge, resources, and community around a thematic area of Data Science, encompassing research, education, and outreach. TAMIDS will support each Lab through a combination of seed funding for new research, effort from TAMIDS personnel, preferred access to existing TAMIDS programs, and organizational and logistical support. TAMIDS is piloting this program through its Scientific Machine Learning Lab (SciML Lab).
Scientific Machine Learning
Scientific Machine Learning (SciML) is a rapidly developing area that aims to revolutionize the practice of Science and Engineering, by bringing together the fields of Machine Learning and Scientific Computation. Typical data-driven Machine Learning methodologies do not incorporate physical understanding of the problem domain. Furthermore, in many scientific domains high-fidelity data are expensive or time-consuming to obtain. Physics-aware SciML addresses both of these problems by introducing regularizing constraints obtained from physical laws, allowing prediction of future performance of complex multiscale, multiphysics systems using sparse, low-fidelity, and heterogeneous data. Unlike traditional black-box Machine Learning methods, SciML aims to deliver interpretable models, leading to improved verification and validation in mission-critical applications.
SciML Lab Goals
There is a great unexplored potential for research and education in Scientific Machine Learning, with many opportunities for high-impact, highly-visible research and extra-mural funding over the next decade. Texas A&M University, with its large research enterprise in engineering, science, and high-performance computing, is uniquely positioned to emerge as one of the national leaders in this area. The purpose of SciML Lab is to help to make this a reality. The lab will foster community building efforts by means of workshops and a seminar series, will develop new educational resources for SciML, including tutorials, hands-on short courses, and a for-credit courses with stacked sections at the undergraduate and graduate levels, all of which will be aimed at audiences drawn from multiple disciplines within Science, Engineering, Geosciences and Agriculture and also promote knowledge of SciML in further disciplines that present opportunities for impact.