Latent ODEs for Irregularly-Sampled Time Series
This addresses a problem in applications with non-uniform time intervals, offering improved modeling for irregularly-sampled data, though it is incremental as it builds on existing Latent ODE frameworks.
The paper tackles modeling irregularly-sampled time series by generalizing RNNs to have continuous-time hidden dynamics defined by ODEs, called ODE-RNNs, and integrating them into Latent ODE models, showing experimentally that these ODE-based models outperform RNN-based counterparts on such data.
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), a model we call ODE-RNNs. Furthermore, we use ODE-RNNs 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 observation times using Poisson processes. We show experimentally that these ODE-based models outperform their RNN-based counterparts on irregularly-sampled data.