![[2510.16161] Still Competitive: Revisiting Recurrent Models for Irregular Time Series Prediction](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.16161] Still Competitive: Revisiting Recurrent Models for Irregular Time Series Prediction
Summary
The paper presents GRUwE, a novel Gated Recurrent Unit model designed for predicting irregularly sampled multivariate time series, demonstrating competitive performance against state-of-the-art methods.
Why It Matters
As irregular time series data becomes increasingly prevalent in fields like healthcare and sensor networks, efficient and effective prediction methods are crucial. GRUwE offers a simpler, computationally efficient alternative to complex architectures, potentially democratizing access to advanced predictive analytics.
Key Takeaways
- GRUwE is a new RNN-based model tailored for irregular time series prediction.
- It utilizes a Markov state representation and two reset mechanisms for improved accuracy.
- Empirical results show GRUwE matches or exceeds the performance of current state-of-the-art models.
- The model is easy to implement and requires minimal hyper-parameter tuning.
- GRUwE significantly reduces computational overhead in online deployments.
Computer Science > Machine Learning arXiv:2510.16161 (cs) [Submitted on 17 Oct 2025 (v1), last revised 18 Feb 2026 (this version, v2)] Title:Still Competitive: Revisiting Recurrent Models for Irregular Time Series Prediction Authors:Ankitkumar Joshi, Milos Hauskrecht View a PDF of the paper titled Still Competitive: Revisiting Recurrent Models for Irregular Time Series Prediction, by Ankitkumar Joshi and 1 other authors View PDF HTML (experimental) Abstract:Modeling irregularly sampled multivariate time series is a persistent challenge in domains like healthcare and sensor networks. While recent works have explored a variety of complex learning architectures to solve the prediction problems for irregularly sampled time series, it remains unclear what the true benefits of some of these architectures are, and whether clever modifications of simpler and more efficient RNN-based algorithms are still competitive, i.e. they are on par with or even superior to these methods. In this work, we propose and study GRUwE: Gated Recurrent Unit with Exponential basis functions, that builds upon RNN-based architectures for observations made at irregular times. GRUwE supports both regression-based and event-based predictions in continuous time. GRUwE works by maintaining a Markov state representation of the time series that updates with the arrival of irregular observations. The Markov state update relies on two reset mechanisms: (i) observation-triggered reset to account for the new obser...
