Corpus ID: 226227575

Interpolation Technique to Speed Up Gradients Propagation in Neural ODEs

@article{Daulbaev2020InterpolationTT,
  title={Interpolation Technique to Speed Up Gradients Propagation in Neural ODEs},
  author={Talgat Daulbaev and A. Katrutsa and L. Markeeva and Julia Gusak and A. Cichocki and I. Oseledets},
  journal={arXiv: Neural and Evolutionary Computing},
  year={2020}
}

• Talgat Daulbaev, A. Katrutsa, +3 authors I. Oseledets
• Published 2020
• Computer Science, Mathematics
• arXiv: Neural and Evolutionary Computing

We propose a simple interpolation-based method for the efficient approximation of gradients in neural ODE models. We compare it with the reverse dynamic method (known in the 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 that was confirmed… Expand

View PDF on arXiv

Save to Library

Create Alert

Cite

Launch Research Feed