Physics-Informed Feature-to-Feature Learning for Design-Space Dimensionality Reduction in Shape Optimisation
Shahroz Khan*, Andrea Serani, Matteo Diez, Panagiotis Kaklis
American Institute of Aeronautics and Astronautics, Scitech 2021 Forum
[Paper] [Presentation] [Video]
This repository contains Matlab implementation of the algorithm framework for Physics-Informed Feature-to-Feature Learning for Dimensionality Reduction, including the implementation of Principal Component Analysis, Active-Subspace Method and Gaussian Process Regression.
Following the different pipelines were tested:
This first pipeline is the Typical Active Subspace Method, second pipeline is the proposed approach, which combaine widely used Principal Compnent Analysis and Active Subspace Method and last pipeline is the Physics-informed Principal Compnent Analysis with Active Subspace Method.
The first author is grateful to the Mac Robertson Trust for sponsoring his visit to CNR-INM through their Postgraduate Travel Scholarship program. CNR-INM authors are grateful to the US Office of Naval Research for its support through NICOP grant N62909-18-1-2033. The first and last author of this work has also received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant "GRAPES: learninG, pRocessing And oPtimising shapES" (agreement No. 860843).