scikit-learn

A new version of scikit-learn is out! And I'm proud that my PR #32100 is among the key highlights of this release 😁 (faster decision trees when using the absolute error). I'd like to thank the reviewers once again for their very thorough reviews! Thanks to them (and thanks to the tests too!), I'm confident that I delivered reliable and maintainable code. The PR: https://lnkd.in/eeEahkv9 The technical report: https://lnkd.in/eWXcSNFv You’ll find a detailed analysis of the different algorithms I considered to address the efficiency challenges of fitting decision trees with the absolute error.

scikit-learn in numbers: 🟠 Downloads: 3.5 billions 🔵 GitHub: 26.5K forked repos; 64.3K stars: https://lnkd.in/dwncfBb7 🟠 Kaggle State of Data Science: scikit-learn consistently ranks as the top machine learning framework 🔵 Monthly website visitors: 1.1 Million unique visitors 🤩 Please consider sponsoring us via GitHub Sponsors ➡️ https://lnkd.in/eYwBG9Yq #GitHub #datascience #machinelearning #opensource

🚀 scikit-learn 1.8 is out 🚀 A big shoutout to the community of contributors who continue to push open-source machine learning forward ❤️ ✨ Key Highlights: ▶️ Expanded Array API support (including PyTorch & CuPy) to run more estimators and metrics on GPUs ▶️ Free-threaded CPython 3.14 support for better multi-threaded performance ▶️ Probability calibration with temperature scaling in CalibratedClassifierCV ▶️ Major efficiency boosts in linear models (Lasso / ElasticNet with gap safe screening) ▶️ Much faster and more robust DecisionTreeRegressor with criterion="absolute_error" ▶️ New manifold.ClassicalMDS implementation for classical multidimensional scaling 🔗 Check the full release highlights: https://lnkd.in/gkXQSbmZ Discover scikit-learn 1.8 and its: 🟢 28 new features 🔵 12 efficiency improvements & 13 enhancements 🟡 9 API changes 🔴 34 fixes 👥 193 contributors (thank you all!) 📖 More details in the release notes: https://lnkd.in/gCrs42se You can upgrade with pip as usual: pip install -U scikit-learn Using conda-forge builds: conda install -c conda-forge scikit-learn #scikitlearn #MachineLearning #opensource #DataScience #Python #ML

📍 The place to be next week - 2 days packed with data science in Paris! We’re excited to bring together the scikit-learn community along with leaders from both technology and business for the very first edition of Probability 1.0, a scikit-learn–centric event designed to share insights, best practices, and real-world impact. A huge thank you to the organiser :probabl. and to our partners Quansight, BNP Paribas, CHANEL, and Inria for making this premiere edition possible. 🎉 Looking forward to two days of learning, collaboration, and open-source excellence.

That’s BIG for Open Source AI and our scikit-learn project! 💥 Congrats to :probabl., the community and its incredible dynamism. This milestone shows how far open source machine learning has come, and how much stronger we are when we build together.

🌱 When open source sparks entrepreneurship: scikit-learn as a launchpad As a community-driven open-source library, scikit-learn has become the foundation for a dynamic ecosystem of entrepreneurs and ventures. ✨ :probabl. created by scikit-learn core developers, helping organizations scale their data science, selling data science expertise while contributing back to open source through skrub, skore, scikit-learn and many more... 🔗 https://probabl.ai and very recently: ✨ Skfolio Labs — co-founded by Hugo Delatte and Daniel Farrell, contributing and supporting Skfolio, an open-source library built on top of scikit-learn, specializing in portfolio optimization and risk management. 🔗 https://skfoliolabs.com 📢 Read the official announcements: Hugo’s post 👉 https://lnkd.in/ePwyU75u Daniel’s post 👉 https://lnkd.in/eqXHjP5r We’re excited to see new projects emerging that extend scikit-learn into specialized domains, showing how open-source innovation can spark sustainable ventures and empower whole industries. #scikitlearn #OpenSource #MachineLearning #DataScience #QuantFinance

A bunch of scikit-learn core contributors will attend or speak at PyData Paris 2025 on Tuesday and Wednesday next week. There is still time to grab a seat to attend the main conference talks and maybe even present a lightning talk ;) Ticketing, practical infos and schedule at: https://lnkd.in/epawxTkH We will also participate in the sprint day on Thursday the 2nd, but unfortunately that day has limited seats and there is already a growing dedicated waitlist. Looking forward to meeting you there, exchanging with experts of all the fields of data science, learning more about your use cases and collecting feedback for the future developments of our library and related projects!

scikit-learn was honored to be selected to participate in Cohort 2 of the GitHub Secure Open Source Fund (OSF) Training Program. It was an intense 3-week training program, with over 90 open source maintainers joining the training. There were numerous workshops delivered by experts at the GitHub Security Lab. ⭐️ For many of these workshops, the learning materials are publicly available, and they are shared in the article here. 🙏Thank you to Gregg Cochran, Kevin Crosby, all the instructors at GitHub Security Lab, and the funders for this pivotal opportunity and support of open source projects. #opensource #security #machinelearning https://lnkd.in/emA_yx6a

The temperature scaling PR scikit-learn/scikit-learn/31068 that Christian Lorentzen and I worked on has finally been merged into scikit-learn’s main branch. Probability calibration with temperature scaling will be available in scikit-learn 1.8. I’d like to thank Olivier Grisel, Christian Lorentzen, and Omar Salman for guiding me through the entire process. For every minute I spent working on the PR, they spent hours reviewing it to ensure algorithmic efficiency, numerical stability, correct unit test implementation, and rigorous documentation. I'd also like to thank David Holzmüller for his hard work in convincing the scikit-learn core maintainers that temperature scaling is a valuable addition to the library --- the merge wouldn’t have been possible without him --- and Adrin J. for his confident vote of support for me to open the PR. https://lnkd.in/gmv2K_gc