[Webinar] Python array API support in scikit-learn for GPU acceleration and TabICLv2

Let’s join us for a deep dive into recent developments in scikit-learn GPU support and scikit-learn compatible tabular foundation models.

Google Colab notebook of the demo: https://colab.research.google.com/drive/1-FiOQK5MTCTHsmv_lBXZKlBc2LjBfBGA?usp=sharing

During this session we will:

• Introduce what is the array API and how it can be used to accelerate scikit-learn models on GPUs,
• Demonstrate how to accelerate a non-linear regression pipeline with a GPU,
• Introduce TabICLv2, a tabular foundation model that can yield state-of-the-art accuracy without hyperparameter tuning,
• Explore different ways to model predictive uncertainty with traditional scikit-learn pipelines and TabICL, especially in the presence of heteroscedasticity,
• Assess the reliability of the predictive uncertainty estimates of different models with both large and small datasets.

We will then conclude with lessons learned from this experiment on the pros and cons of each approach in small and large dataset settings and take questions from the audience.

This webinar was recorded on the 26th of March, 2026.