/burgers

Neural solver to compute solutions to Burger's equation. 1D + 3D.

Primary LanguageJupyter NotebookMIT LicenseMIT

Burger's equation DL solver

Physics differential equations are a great opportunity to play with DL frameworks' auto-differentiation features and develop a poorly ambitious solver from scratch. On the pros side it's fun, interpolation comes for free, and it's a great pretext for practicing the frameworks — the cons are it's sillily unefficient and practically useless. Still enjoyable though.

  • 1d for 1D case. Actually usable.
  • 3d for 3D case. Don't use this. It will generate ~35PiB of data with just n = 100 and smoke your hardware for a laugh.