LiCROM: Linear-subspace continuous reduced order modeling with neural fields

Abstract

We propose LiCROM, a method that combines linear-subspace model reduction with continuous neural field representations. By representing the reduced basis using neural fields, we enable continuous queries in space and time, facilitating super-resolution reconstruction and smooth interpolation. This approach bridges classical model reduction techniques with modern neural representations, offering advantages for computer graphics applications including real-time physics simulation and animation.

Publication
SIGGRAPH Asia 2023 Conference Papers, pages 1–12, 2023

Related