Projection-based multifidelity linear regression for data-scarce applications
The episode discusses a paper on projection-based multifidelity linear regression designed for data-scarce engineering applications. The authors propose combining limited high-fidelity simulation data with abundant low-fidelity data using data augmentation and weighted least squares regression. This method achieves a 3% to 12% accuracy improvement over single-fidelity models without increasing computational cost, offering greater robustness.