Skip to content

Home

Documentation page for the Deep learning based Model Discovery package DeepMoD. DeePyMoD is a PyTorch-based implementation of the DeepMoD algorithm for model discovery of PDEs and ODEs.github.com/PhIMaL/DeePyMoD. This work is based on two papers: The original DeepMoD paper arXiv:1904.09406, presenting the foundation of this neural network driven model discovery and a follow-up paper arXiv:2011.04336 describing a modular plug and play framework.

Summary

Screenshot DeepMoD is a modular model discovery framewrok aimed at discovering the ODE/PDE underlying a spatio-temporal dataset. Essentially the framework is comprised of four components:

  • Function approximator, e.g. a neural network to represent the dataset,
  • Function library on which the model discovery is performed,
  • Constraint function that constrains the neural network with the obtained solution
  • Sparsity selection algorithm.