Neural Flows: Efficient Alternative to Neural ODEs

Marin Biloš, Johanna Sommer, Syama Sundar Rangapuram, Tim Januschowski, Stephan Günnemann

Keywords: neural flow, neural ode, time series, density estimation, normalizing flow

Abstract: Neural ordinary differential equations describe how values change in time. This is the reason why they gained importance in modeling sequential data, especially when the observations are made at irregular intervals. In this paper we propose an alternative by directly modeling the solution curves - the flow of an ODE - with a neural network. This immediately eliminates the need for expensive numerical solvers while still maintaining the modeling capability of neural ODEs. We propose several flow architectures suitable for different applications by establishing precise conditions on when a function defines a valid flow. Apart from computational efficiency, we also provide empirical evidence of favorable generalization performance via applications in time series modeling, forecasting, and density estimation.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

TL;DR: We directly model the ODE solutions, which is much faster and achieves better results on time series applications.

Code: https://www.daml.in.tum.de/neural-flows

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2110.13040/code)