Interpolation Technique to Speed up Gradients Propagation in Neural ODEs

Abstract

We propose a simple interpolation-based method for the efficient approximation of gradients in neural ODE models. We compare it with reverse dynamic method (known in literature as “adjoint method”) to train neural ODEs on classification, density estimation and inference approximation tasks. We also propose a theoretical justification of our approach using logarithmic norm formalism. As a result, our method allows faster model training than the reverse dynamic method what was confirmed and validated by extensive numerical experiments for several standard benchmarks.

PDF NeurIPS Semantic Scholar

Cite

Text

Daulbaev et al. "Interpolation Technique to Speed up Gradients Propagation in Neural ODEs." Neural Information Processing Systems, 2020.

Markdown

[Daulbaev et al. "Interpolation Technique to Speed up Gradients Propagation in Neural ODEs." Neural Information Processing Systems, 2020.](https://mlanthology.org/neurips/2020/daulbaev2020neurips-interpolation/)

BibTeX

@inproceedings{daulbaev2020neurips-interpolation,
  title     = {{Interpolation Technique to Speed up Gradients Propagation in Neural ODEs}},
  author    = {Daulbaev, Talgat and Katrutsa, Alexandr and Markeeva, Larisa and Gusak, Julia and Cichocki, Andrzej and Oseledets, Ivan},
  booktitle = {Neural Information Processing Systems},
  year      = {2020},
  url       = {https://mlanthology.org/neurips/2020/daulbaev2020neurips-interpolation/}
}