Lu Lu’s innovative approach to data science is attracting attention from diverse fields such as biology, physics, and even the insurance industry. His methodologies, which incorporate artificial intelligence (AI) for fundamental scientific research, have piqued the interest of researchers and organizations worldwide.
Lu, an assistant professor of statistics and data science at Yale University, has emerged as a key figure in “Physics-Informed Machine Learning.” This approach integrates traditional data science techniques with physical principles and equations, offering new ways to understand complex systems.
“I’m not a domain expert in geoscience, biology, fire insurance, climate change, heart or blood diseases, or astronomy,” Lu explained. “But I have tools that can help in all of those areas.”
Over the past few years, this concept has gained traction as more researchers explore its potential to advance scientific discovery. Lu’s work, supported by various grants, is breaking new ground in several high-impact areas.
Data Science Meets Physics
Lu Lu: The traditional approach to scientific computing involves using computational techniques to simulate systems with known mathematical equations. However, in fields like biology, fully understanding a system before simulation is often impractical. Machine learning usually requires extensive data, which is not always available. Our method bridges this gap by combining limited data sets with physics and machine learning.
This innovative method started with a small team experimenting with integrating partial differential equations and machine learning over five years ago. Despite initial skepticism, it has gained recognition as a valuable research tool.
Applications Across Disciplines
One of Lu’s projects, supported by a $4 million grant from the U.S. Department of Energy (DOE), focuses on developing federated AI methods to handle large data sets while preserving privacy. This approach is particularly useful for analyzing climate data collected by multiple institutions without the need for extensive data transfers.
Another exciting application of Lu’s work is in geoscience, where AI methods are used to study Earth’s subsurface structure, with implications for earthquake research and energy exploration.
Advancing Biological Research
Lu’s research, funded by the National Science Foundation, also extends to biological sciences. The project aims to understand genomic organization in cell nuclei, using sparse and noisy data to conduct robust research over three years.
Collaborations with Industry
Lu’s collaboration with FM Global, a major commercial property insurance company, involves simulating fire damage to buildings using AI. Traditional modeling techniques in the insurance industry are costly, but preliminary results from Lu’s team show promise in offering a more efficient solution.
Impacting Science and Society
Lu finds it exhilarating to apply his methods across various scientific and societal domains. He recently co-authored a paper in Nature Computational Science on using AI to model the human heart’s geometry. This advancement could significantly benefit patients by reducing the time doctors spend creating models for each individual.
Lu’s work continues to inspire and push the boundaries of what is possible with AI in science.
Original Story at news.yale.edu