Paper-Conference

SMART: Scalable Mesh-free Aerodynamic Simulations from Raw Geometries using a Transformer-based Surrogate Model
We introduce SMART, a neural surrogate model for 3D aerodynamic simulations that does not require access to the expensive simulation mesh during inference.
Jan Hagnberger, Mathias Niepert
PDF Code
CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs
We introduce CALM-PDE, a model based on continuous convolution for discretization-agnostic solving of time-dependent partial differential equations, which achieves strong performance in fluid dynamics simulations.
Jan Hagnberger, Daniel Musekamp, Mathias Niepert
PDF Code
Vectorized Conditional Neural Fields: A Framework for Solving Time-dependent Parametric Partial Differential Equations
We propose a continuous model based on conditional neural fields and linear transformers for solving time-dependent parametric partial differential equations.
Jan Hagnberger, Marimuthu Kalimuthu, Daniel Musekamp, Mathias Niepert
PDF Code Project