Home
About
Mission
Affiliates
Computing Infrastructure
People
Research
Computational Cardiovascular Biomechanics
Physics-Informed Machine Learning for the Solution of PDEs
Railway Dynamic Sytem Modelling with Physics-Informed Neural Network
Computational Fluid Dynamics
Computational Linear Algebra
Parallel Computing
Projects
TUBITAK – (Scientific and Technological Research Council of Türkiye) Funded Projects
International Projects
ROKETSAN – (Defense Company of Turkey) Funded Projects
ITU Research Fund (BAP) Supported Projects
Publications
News
Events
Join Us
Contact
Home
/
News Detail
/
HiPC 2025: 32nd IEEE International Conference on High Performance Computing, Data, and Analytics - Afrah Farea
Title:
Multi-Objective Loss Balancing in Physics-Informed Neural Networks for Fluid Flow Applications
At the 32nd IEEE International Conference on High Performance Computing, Data, and Analytics, held in India, Afrah Farea delivered a presentation on improving the accuracy of Physics-Informed Neural Networks (PINNs) for fluid flow applications.
Paper Abstract
Physics-Informed Neural Networks (PINNs) have emerged as a promising machine learning approach for solving partial differential equations (PDEs). However, PINNs face significant challenges in balancing multi-objective losses, as multiple competing loss terms such as physics residuals, boundary conditions, and initial conditions must be appropriately weighted. While various loss balancing schemes have been proposed, they have been implemented within neural network architectures with fixed activation functions, and their effectiveness has been assessed using simpler PDEs. We hypothesize that the effectiveness of loss balancing schemes depends not only on the balancing strategy itself, but also on the loss function design and the neural network's inherent function approximation capabilities, which are influenced by the choice of activation function. In this paper, we extend existing solutions by incorporating trainable activation functions within the neural network architecture and evaluate the proposed approach on complex fluid flow applications modeled by the Navier-Stokes equations. Our evaluation across diverse Navier-Stokes problems demonstrates that this proposed solution achieves root mean square error (RMSE) improvements ranging from 7.4% to 95.2% across different scenarios. These findings highlight the importance of carefully designing the loss function and selecting activation functions for effective loss balancing.
The publication can be accessed via the link:
https://doi.org/10.48550/arXiv.2509.14437
Date
09 Jan 2026
Share