Войти
аналитика - Neural PDE Solvers
Latent Neural PDE Solver: Accelerating Simulation of Partial Differential Equations
The author proposes a Latent Neural PDE Solver framework to accelerate the simulation of systems governed by partial differential equations by operating in a mesh-reduced space, achieving competitive accuracy and efficiency compared to traditional methods.
SineNet: A Multi-Stage Architecture for Learning Temporal Dynamics in Time-Dependent Partial Differential Equations
SineNet is a multi-stage neural network architecture that effectively models the temporal dynamics in time-dependent partial differential equations by reducing the misalignment in skip connections between the downsampling and upsampling paths.