TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Latent ODEs for Irregularly-Sampled Time Series

Latent ODEs for Irregularly-Sampled Time Series

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

2019-07-08Multivariate Time Series ImputationTime SeriesMultivariate Time Series ForecastingTime Series AnalysisTime Series Classification
PaperPDFCodeCodeCodeCodeCodeCodeCodeCodeCode(official)CodeCode

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), 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.

Results

TaskDatasetMetricValueModel
ImputationPhysioNet Challenge 2012mse (10^-3)2.118Latent ODE (ODE enc)
ImputationPhysioNet Challenge 2012mse (10^-3)2.789Latent ODE + Poisson
ImputationMuJoCoMSE (10^2, 50% missing)0.285Latent ODE (ODE enc)
ImputationMuJoCoMSE (10^2, 50% missing)0.665ODE-RNN
Time Series ForecastingMuJoCoMSE (10^-2, 50% missing)1.258Latent ODE (ODE enc)
Time Series ForecastingMuJoCoMSE (10^-2, 50% missing)26.463ODE-RNN
Time Series ForecastingPhysioNet Challenge 2012MSE stdev0.05Latent ODE + Poisson
Time Series ForecastingPhysioNet Challenge 2012mse (10^-3)2.208Latent ODE + Poisson
Time Series ForecastingPhysioNet Challenge 2012MSE stdev0.029Latent ODE (ODE enc)
Time Series ForecastingPhysioNet Challenge 2012mse (10^-3)2.231Latent ODE (ODE enc)
Feature EngineeringPhysioNet Challenge 2012mse (10^-3)2.118Latent ODE (ODE enc)
Feature EngineeringPhysioNet Challenge 2012mse (10^-3)2.789Latent ODE + Poisson
Feature EngineeringMuJoCoMSE (10^2, 50% missing)0.285Latent ODE (ODE enc)
Feature EngineeringMuJoCoMSE (10^2, 50% missing)0.665ODE-RNN
Time Series AnalysisMuJoCoMSE (10^-2, 50% missing)1.258Latent ODE (ODE enc)
Time Series AnalysisMuJoCoMSE (10^-2, 50% missing)26.463ODE-RNN
Time Series AnalysisPhysioNet Challenge 2012MSE stdev0.05Latent ODE + Poisson
Time Series AnalysisPhysioNet Challenge 2012mse (10^-3)2.208Latent ODE + Poisson
Time Series AnalysisPhysioNet Challenge 2012MSE stdev0.029Latent ODE (ODE enc)
Time Series AnalysisPhysioNet Challenge 2012mse (10^-3)2.231Latent ODE (ODE enc)
Multivariate Time Series ForecastingMuJoCoMSE (10^-2, 50% missing)1.258Latent ODE (ODE enc)
Multivariate Time Series ForecastingMuJoCoMSE (10^-2, 50% missing)26.463ODE-RNN
Multivariate Time Series ForecastingPhysioNet Challenge 2012MSE stdev0.05Latent ODE + Poisson
Multivariate Time Series ForecastingPhysioNet Challenge 2012mse (10^-3)2.208Latent ODE + Poisson
Multivariate Time Series ForecastingPhysioNet Challenge 2012MSE stdev0.029Latent ODE (ODE enc)
Multivariate Time Series ForecastingPhysioNet Challenge 2012mse (10^-3)2.231Latent ODE (ODE enc)

Related Papers

MoTM: Towards a Foundation Model for Time Series Imputation based on Continuous Modeling2025-07-17The Power of Architecture: Deep Dive into Transformer Architectures for Long-Term Time Series Forecasting2025-07-17Emergence of Functionally Differentiated Structures via Mutual Information Optimization in Recurrent Neural Networks2025-07-17Data Augmentation in Time Series Forecasting through Inverted Framework2025-07-15D3FL: Data Distribution and Detrending for Robust Federated Learning in Non-linear Time-series Data2025-07-15Towards Interpretable Time Series Foundation Models2025-07-10MoFE-Time: Mixture of Frequency Domain Experts for Time-Series Forecasting Models2025-07-09Foundation models for time series forecasting: Application in conformal prediction2025-07-09