TACD-GRU: Time-Aware Context-Dependent Autoregressive Model for Irregularly Sampled Time Series
Keywords: Time series models, Irregularly sampled time-series, Autoregressive models, Recurrent neural networks
Abstract: Multivariate time series data and their models are extremely important for under-
standing the behavior of various natural and man-made systems. Development of
accurate time series models often requires capturing intricate relationships among
the variables and their dynamics. Particularly challenging to model and learn
are time series with irregular and sparse observations, that may arise in domains
as diverse as healthcare, sensor and communication networks. In this work, we
propose and study TACD-GRU, a new Time- Aware Context-Dependent Gated
Recurrent Unit framework for multivariate time series prediction (or forecasting)
that accounts for irregularities in observation times of individual time series vari-
ables and their dependencies. Our framework defines a novel sequential unit that
is triggered by the arrival of a new observation to update its state, and a predic-
tion module that supports time series predictions at any future time. The current
prediction module consists of and combines two novel prediction models: (i) a
context-based model (TACD-GRU-CONTEXT) that relies on a set of tunable latent
decay functions of time and their linear combinations to support the prediction,
and (ii) an attention-based model (TACD-GRU-ATTENTION) that models depen-
dencies among variables and their most recent values using a temporal attention
mechanism. Our model shows highly competitive performance when powered by
both individual and combined prediction functions outperforming existing state-of-
the-art (SOTA) models on both single-step and multi-step prediction tasks across
three real-world datasets.
Supplementary Material: zip
Primary Area: learning on time series and dynamical systems
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13435
Loading