Figure 1. Happy ML researcher and open source developer presenting his toolbox SKADA at PyData Paris. Congrats @tgnassou.bsky.social the presentation was awesome!

📩 Message me if you’d like to participate!

Skada: github.com/scikit-adapt...

🎯 Goal of the Skada Coding Sprint
- Improve Skada: add new methods, improve documentation, fix bugs ...
- Contribute to open source in a welcoming environment
- Collaborate with a community of ML researchers and developers
- Implement and test your own Domain Adaptation methods
- Have a lot of fun!

💡 To make it easier for everyone to apply these techniques, we built Skada: a simple, Python-based library for Domain Adaptation.

Skada team organizes a coding sprint :(@rflamary.bsky.social , @antoinecollas.bsky.social , @ambroiseodt.bsky.social)

📍 When: June 24–25
📍 Where: Inria Saclay

Skada Sprint Alert: Contribute to Domain Adaptation in Python

📖 Machine learning models often fail when the data distribution changes between training and testing. That’s where Domain Adaptation comes in — helping models stay reliable across domains.

SKADA-Bench is built on SKADA: github.com/scikit-adapt...

This work results from a collaboration with Yanis Lalou, @antoinecollas.bsky.social , Antoine de Mathelin, Oleksii Kachaiev, @ambroiseodt.bsky.social , Alexandre Gramfort, Thomas Moreau and @rflamary.bsky.social !

The benchmark shows deep DA methods struggle beyond computer vision, highlighting their limits on other modalities!

The results show the benefit of DA in some cases but parameter-sensitive shallow methods struggle to adapt to new domains. Better to use low-parameter methods like LinOT & Coral!

This benchmark is done using a realistic scenario comprising the validation of hyperparameters using nested loop and DA scorers!

🔬 What’s inside?
• Multi-Modality Benchmark: 4 simulated + 8 real datasets
• 20 Shallow DA Methods: Reweighting, mapping, subspace alignment & others
• 7 Deep DA Methods: CAN, MCC, MDD, SPA & more
• 7 Unsupervised Validation Scorers

DA adapts machine learning models to distribution shifts between training and test sets. We propose SKADA-Bench, the first comprehensive, reproducible benchmark that evaluates DA methods across multiple modalities: computer vision, natural language processing, tabular, and biomedical data.

🚀 I’m pleased to announce a new preprint!

"SKADA-Bench: Benchmarking Unsupervised Domain Adaptation Methods with Realistic Validation On Diverse Modalities"

📢 Check it out & contribute!
📜 Paper: arxiv.org/abs/2407.11676
💻 Code: github.com/scikit-adapt...

This library is a team effort: @antoinecollas.bsky.social, Oleksii Kachaiev, @rflamary.bsky.social , Yanis Lalou, Antoine de Mathelin, Ruben Bueno, Apolline Mellot, @ambroiseodt.bsky.social , Alexandre Gramfort and myself!

🔀 Improved Subsampling Tools
- Added StratifiedDomainSubsampler and DomainSubsampler to handle large datasets effortlessly.

🤖 Deep Model Enhancements
- Smarter batch handling.
- Many bug fixes.

📖 Documentation Upgrades
- Contributor Guide: Join the development of Skada!
- New Logo!!

📊 Advanced Scorers
- New MixValScorer for mixup validation.
- Enhanced scorer compatibility with deep models.

💡New Deep Domain Adaptation Methods: CAN, SPA, MCC, and MDD.These methods combine the cross entropy loss on the source domain with domain aware losses (graph based, adversarial, class confusion, …).

💡New Shallow Domain Adaptation Methods: MongeAlignment and JCPOT for linear multi-source domain adaptation with optimal transport.

🚀 Skada v0.4.0 is out!

Skada is an open-source Python library built for domain adaptation (DA), helping machine learning models to adapt to distribution shifts.
Github: github.com/scikit-adapt...
Doc: scikit-adaptation.github.io
DOI: doi.org/10.5281/zeno...
Installation: `pip install skada`