Latent Ordinary Differential Equations for Irregularly-Sampled Time Series

Yulia Rubanova, Ricky T. Q. Chen, David Duvenaud

Abstract: Time series with non-uniform intervals occur in many applications, and are difficult to model using standard recurrent neural networks (RNNs). We generalize RNNs to have continuous-time hidden dynamics defined by ordinary differential equations (ODEs), which we call ODE-RNNs. We use ODE-RNN to replace the recognition network of the recently-proposed Latent ODE model. Both ODE-RNNs and Latent ODEs can naturally handle arbitrary time gaps between observations, and can explicitly model the probability of different observation times using Poisson processes. We show experimentally that these ODE-based models outperform RNN-based counterparts on irregularly-sampled data.

Code Link: https://github.com/YuliaRubanova/latent_ode/

CMT Num: 2858