Learning from Irregularly-Sampled Time Series: A Missing Data\n Perspective

2020/08/17 by Steven Cheng-Xian Li, Li, Steven Cheng-Xian, Benjamin M. Marlin +1 · 10 citations
Computer Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning in Healthcare #Time Series Analysis and Forecasting

paper · pdf · doi:10.48550/arxiv.2008.07599

openalex publication_date 2020/08/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Irregularly-sampled time series occur in many domains including healthcare.\nThey can be challenging to model because they do not naturally yield a\nfixed-dimensional representation as required by many standard machine learning\nmodels. In this paper, we consider irregular sampling from the perspective of\nmissing data. We model observed irregularly-sampled time series data as a\nsequence of index-value pairs sampled from a continuous but unobserved\nfunction. We introduce an encoder-decoder framework for learning from such\ngeneric indexed sequences. We propose learning methods for this framework based\non variational autoencoders and generative adversarial networks. For continuous\nirregularly-sampled time series, we introduce continuous convolutional layers\nthat can efficiently interface with existing neural network architectures.\nExperiments show that our models are able to achieve competitive or better\nclassification results on irregularly-sampled multivariate time series compared\nto recent RNN models while offering significantly faster training times.\n

Cited by

• Dynamical modeling of nonlinear latent factors in multiscale neural activity with real-time inference
• Towards Stable and Structured Time Series Generation with Perturbation-Aware Flow Matching
• Prepared for the Unknown: Adapting AIOps Capacity Forecasting Models to Data Changes
• ASTGI: Adaptive Spatio-Temporal Graph Interactions for Irregular Multivariate Time Series Forecasting
• Incomplete Data, Complete Dynamics: A Diffusion Approach
• MTM: A Multi-Scale Token Mixing Transformer for Irregular Multivariate Time Series Classification
• Neural Ordinary Differential Equations for Learning and Extrapolating System Dynamics Across Bifurcations
• Causal Discovery for Irregularly Time Series with Consistency Guarantees
• Online Coreset Selection for Learning Dynamic Systems
• Lung Cancer Risk Estimation with Incomplete Data: A Joint Missing Imputation Perspective

Related

• Time-Parameterized Convolutional Neural Networks for Irregularly Sampled Time Series
• A Survey on Principles, Models and Methods for Learning from Irregularly Sampled Time Series
• Improving Irregularly Sampled Time Series Learning with Dense\n Descriptors of Time
• Multi-Time Attention Networks for Irregularly Sampled Time Series
• Finding Short Signals in Long Irregular Time Series with Continuous-Time Attention Policy Networks