PINN training using biobjective optimization: the trade-off between data loss and residual loss
DOI10.1016/j.jcp.2023.112211arXiv2302.01810MaRDI QIDQ6162878
Fabian Heldmann, Kathrin Klamroth, Matthias Ehrhardt, Sarah Berkhahn
Publication date: 16 June 2023
Published in: Journal of Computational Physics (Search for Journal in Brave)
Full work available at URL: https://arxiv.org/abs/2302.01810
multiobjective optimizationloss functionPareto frontweighting parameterscompartment modelsphysics-informed neural networks
Mathematical programming (90Cxx) Artificial intelligence (68Txx) Numerical methods for partial differential equations, initial value and time-dependent initial-boundary value problems (65Mxx)
Cites Work
- Unnamed Item
- Multiple-gradient descent algorithm (MGDA) for multiobjective optimization
- Steepest descent methods for multicriteria optimization.
- Estimating the time-dependent contact rate of SIR and SEIR models in mathematical epidemiology using physics-informed neural networks
- NSFnets (Navier-Stokes flow nets): physics-informed neural networks for the incompressible Navier-Stokes equations
- A nonstandard finite difference scheme for the SVICDR model to predict COVID-19 dynamics
- Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems
- Meta-learning PINN loss functions
- Scientific machine learning through physics-informed neural networks: where we are and what's next
- Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
- Efficient and sparse neural networks by pruning weights in a multiobjective learning approach
- Integrating Approximation and Interactive Decision Making in Multicriteria Optimization
- Exact solutions to a finite-difference model of a nonlinear reaction-advection equation: Implications for numerical analysis
- Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks
- fPINNs: Fractional Physics-Informed Neural Networks
- An Introduction to Mathematical Epidemiology
- Multicriteria Optimization
This page was built for publication: PINN training using biobjective optimization: the trade-off between data loss and residual loss
