Universal Time-Series Representation Learning: A Survey
Universal Time-Series Representation Learning: A Survey
Authors: Patara Trirat, Yooju Shin, Jun-Hyeok Kang, Youngeun Nam, Jihye Na, Minyoung Bae, Jo‐Eun Kim, Byunghyun Kim, Jae-Gil Lee Date: 2026-05-22 Paper ID: openalex:2401.03717
Summary
This survey provides a structured overview of the landscape of universal time-series representation learning. It proposes a novel taxonomy categorized by three fundamental design elements to classify and evaluate existing deep learning approaches. By synthesizing current research, the paper offers insights into effective representation extraction strategies and outlines critical future research directions for the community.
Key Contributions
- Introduces a novel taxonomy for universal time-series representation learning based on three core design elements.
- Provides a comprehensive review and comparative analysis of existing deep learning-based representation learning methods.
- Identifies and discusses common experimental configurations and provides future research directions to guide the field.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2401.03717
- paper_id
- 2401.03717
- paper_source
- openalex
- domain
- nlp
- tags
-
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-05-25T06:32:30Z