Skip to main content

Time Series Classification with TiRex

Paper

Classification is one of the core capabilities of TiRex, leveraging the power of xLSTM to extract meaningful features from time series data efficiently — even across highly irregular or complex time series.

This section provides a structured overview of how TiRex approaches classification from both theoretical and practical perspectives.

📘 Theory​

Learn the fundamental ideas behind classification with TiRex. Includes explanations of:

  • Zero-shot protocol with frozen TiRex forecasting model as feature extractor
  • Embedding extraction and aggregation strategies
  • Multivariate time series handling
  • Embedding augmentation techniques
  • Conceptual summary of the classification transformation pipeline

🧭 Workflow​

Step-by-step playbook that bridges the paper and the tutorials:

  • Data preparation
  • Model initialization
  • Training and evaluation workflows
  • Troubleshooting checklist for common issues

⚙️ Practice​

Dive into practical classification with TiRex. Hands-on tutorials and examples showing:

  • Installation and setup of TiRex classification
  • Data preprocessing
  • Initializing and configuring classifiers
  • Fitting models to your data
  • Making predictions and evaluating results

Together, these guides form a complete introduction to applying TiRex for real-world time series classification.

Cite​

If you use TiRex for Time Series Classification in your research, please cite our work:

@inproceedings{auer:25tirexclassification,
    title = {Pre-trained Forecasting Models: Strong Zero-Shot Feature Extractors for Time Series Classification},
    author = {Andreas Auer and Daniel Klotz and Sebastinan B{\"o}ck and Sepp Hochreiter},
    booktitle = {NeurIPS 2025 Workshop on Recent Advances in Time Series Foundation Models (BERT2S)},
    year = {2025},
    url = {https://arxiv.org/abs/2510.26777},
}