Physics-Enhanced Machine Learning

Physics-Enhanced Machine Learning

With the rapid development in measurement technology and computing power, our interest lies in learning models that explain underlying dynamical behaviors and draw meaningful conclusions from the learned model. The PML team is particularly dedicated to developing innovative machine learning methodologies by incorporating fundamental principles and empirical knowledge of the process. Moreover, we focus on inferring low-dimensional models using machine learning tools, thus reducing computational cost of high-fidelity models.

Projects

Partners: MPI Düsseldorf, Germany
Funded by:
  BiGMax
Funding period: Since 2021
Contact: Vasileios Athanasiou, Peter Benner, Pawan Goyal more

Operator Inference for Learning Reduced-Order Models

Partners: UT Austin USA, CIM NY USA, UCSD USA, METU Turkey
Funded by:
  MPI
Funding period: Since 2019
Contact: Pawan Goyal, Peter Benner

Inferring Dynamical Systems using Frequency Response Data

Partners: UCL Louvain
Funded by:
  MPI
Funding period: Since 2020
Contact: Pawan Goyal, Igor Pontes Duff Pereira, Peter Benner

Physics-Enhanced Deep Learning-Based Image Reconstruction

Funded by:  MPI
Funding period: Since 2020
Contact: Pawan Goyal, Peter Benner

Feature Publications

Benner, P.; Goyal, P. K.; Heiland, J.; Pontes Duff, I.: Operator Inference and Physics-Informed Learning of Low-Dimensional Models for Incompressible Flows. Electronic Transactions on Numerical Analysis: Special Issue SciML (accepted)
Benner, P.; Goyal, P. K.; Kramer, B.; Peherstorfer, B.; Willcox, K.: Operator Inference for Non-Intrusive Model Reduction of Systems with Non-Polynomial Nonlinear Terms. Computer Methods in Applied Mechanics and Engineering 372, 113433, 17 pages (2020)
Benner, P.; Goyal, P. K.; Van Dooren, P.: Identification of Port-Hamiltonian Systems from Frequency Response Data. Systems & Control Letters 143, 104741, 9 pages (2020)
Go to Editor View