• Corpus ID: 245906022

Generative time series models using Neural ODE in Variational Autoencoders

@article{Garsdal2022GenerativeTS,
  title={Generative time series models using Neural ODE in Variational Autoencoders},
  author={M. L. Garsdal and V. Sogaard and S. M. Sorensen},
  journal={ArXiv},
  year={2022},
  volume={abs/2201.04630}
}
In this paper we present an extension to the work done in [1] by implementing Neural Ordinary Differential Equations in a Variational Autoencoder setting for generative time series modeling. An object-oriented approach to the code was taken to allow for easier development and research and all code used in the paper can be found in this Github link. The results in [1] were initially recreated and the reconstructions compared to a baseline LongShort Term Memory AutoEncoder. The model was then… 
A dynamical systems based framework for dimension reduction
TLDR
A novel framework for learning a low-dimensional representation of data based on nonlinear dynamical systems, which is called dynamical dimension reduction (DDR), and it is proved that the resulting optimization problem is well-posed and several properties of the DDR method are established.

References

Extend the scope of the research by testing more specific NODE model setups. • Perform an analysis varying the ODE topology and investigating the impact on model performance
  • 2019