Skip to main content

Time Series Regression with TiRex

Regression is a key capability of TiRex, leveraging the power of xLSTM to extract meaningful features from time series data efficiently — even across highly irregular or complex time series.

TiRex regression uses the same approach as classification: a frozen pre-trained TiRex forecasting model extracts embeddings, which are then fed into a regression head to predict continuous target values.

This section provides a structured overview of how to apply TiRex for regression tasks.

📘 Theory​

For detailed theoretical understanding, please refer to the Classification Theory page, which covers:

  • Zero-shot protocol with frozen TiRex forecasting model as feature extractor
  • Embedding extraction and aggregation strategies
  • Multivariate time series handling
  • Embedding augmentation techniques

The only difference is the task-specific head: regression head maps embeddings to continuous target values.

🧭 Workflow​

Step-by-step guide for applying TiRex to regression tasks:

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

⚙️ Practice​

Hands-on tutorials and examples showing:

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

Together, these guides provide a complete introduction to applying TiRex for real-world time series regression tasks.