Case Study: Stanford University achieves improved predictive model accuracy and generalizability with MathWorks' MATLAB and Deep Learning Toolbox

Using Physics-Informed Machine Learning to Improve Predictive Model Accuracy

Stanford University’s Department of Bioengineering, led by Dr. Sam Raymond, faced the problem that conventional machine learning models for climate and fluid systems often neglect underlying physics—limiting extrapolation—and that incorporating physics via PDEs and boundary conditions is computationally intensive. To tackle this, Stanford partnered with MathWorks and used MATLAB and Deep Learning Toolbox to bring scientific computing and deep learning together.

MathWorks’ Deep Learning Toolbox and MATLAB enabled a physics‑informed machine learning workflow that integrates computational fluid dynamics and field data with deep learning and high‑performance computing. The solution improved predictive model accuracy, efficiency, and generalizability for climate modeling, provided a cohesive framework for signal analysis, image processing, HPC, and deep learning, and eliminated the need to know the exact governing equations to perform the analysis.

Stanford University

Sam Raymond

Post-Doctoral Fellow, Department of Bioengineering