Double-Stitch: Simulation-Free Learning of Population Dynamics
What it is
This paper introduces Double-Stitch, a method for learning Lagrangian mechanics in Wasserstein space without running numerical solvers during training. It works by penalizing the residual of the equation of motion derived from a Clebsch variational principle along a learned population path, enabling faster training than simulation-based approaches.
The problem it solves
Existing methods for learning Lagrangian mechanics in Wasserstein space are simulation-based and expensive, requiring a numerical solver at every training step.
What you could use it for
- Modeling the evolution of cell populations from unpaired snapshots
- Extrapolating fluid dynamics and ocean vortex behavior
- Learning conservative or periodic dynamics that gradient flows cannot describe
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.