🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

skore has a ⭐ CrossValidationReport ⭐ that provides you with an EstimatorReport for each fold of your cross-validation, enabling you to inspect your estimator on each fold. 📽️ YouTube video demo (~3 min): eu1.hubs.ly/H0gCMtf0 🔗 Skore GitHub repo: eu1.hubs.ly/H0gCLyv0

There is a subtle, but important, difference between metrics and scorers in scikit-learn. Understanding this difference really matters, especially when you want to apply ML in business with custom metrics. So we made a new tutorial on the topic: eu1.hubs.ly/H0gFd9Y0

Metrics matter, but sometimes ... model calibration matters more. And there is a small risk that metrics hide that fact. Our latest YT video explains why. eu1.hubs.ly/H0gQ0qN0

skore v0.7.1 is out! We have a new ⭐ ComparisonReport ⭐ that enables you to compare several skore estimator reports on the same test data, as in a benchmark of estimators. 📽️ YouTube video demo (~3 min): eu1.hubs.ly/H0h0YNg0 🔗 Skore GitHub repo: eu1.hubs.ly/H0h0NDn0

Suppose that you know that you have a badly calibrated model, can you fix it? Thanks to some techniques in scikit-learn: yes! Our latest whiteboard video gives all the details: eu1.hubs.ly/H0gQ3tn0

It might sound strange when you hear it first, but in *most* applications ... accuracy rarely matters. It's because the outcome for the business usually depends on something that's quantifiably different. Our latest whiteboard explains the details: eu1.hubs.ly/H0gP_w-0

Metadata routing is a recent addition to the scikit-learn library that will allow for a lot more advanced pipelines and machine learning systems. Our latest YouTube video gives you all the details: eu1.hubs.ly/H0gP_sf0

Tuning a threshold of a classifier tends to have a big impact, maybe even more impact than tuning a hyperparameter. And as of recently, you can even automate this tuning in scikit-learn. Our YouTube channel gives a demo: eu1.hubs.ly/H0gQ6mW0

It's possible that you've been building classifiers for years without ever having heard of variable thresholds. But oh my! They make an existing classifier ✨better✨! Our latest whiteboard video explains why it's one of our most favourite techniques. eu1.hubs.ly/H0gQ7v30

⌛ Enriching our metrics! In the new version 0.8.2 of skore, we added some metrics: you can now access to fit (on the train set only, obviously) and predict time (on whatever dataset you want). To know more about it, you can check the docs: eu1.hubs.ly/H0j4Qbn0

Friday is a great day to take a step back and learn new skills, isn't it? 📖 To learn more about the difference between metrics & scorers in scikit-learn and to use them more easily, check out with short and instructive video: eu1.hubs.ly/H0jypwL0

New in skore 0.8.2! You can now access directly to the prediction inside your reports. Still in the direction to be all included, easily accessible and stored together. ⭐ To be posted of skore's releases, you can star the repo: eu1.hubs.ly/H0jJSFP0

🎉 We are very happy to celebrate our 11th external contributor on skore! Check out: 👉 the repo: eu1.hubs.ly/H0jJSC-0 ⭐ 👉 or the docs: eu1.hubs.ly/H0jJVb80 📑

Last week, we hosted an exciting challenge bringing together data scientists for an exciting team-building event! Participants predicted bike counter data in Paris, diving into real-world data science with tools like skore and skrub. Check out skore: eu1.hubs.ly/H0j-zxV0

🎤 Next week, our product engineer Marie Sacksick will be presenting how to extend scikit-learn with skore, but also with skrub and skops. Thanks Pyladies Paris for this opportunity! To book your seat: eu1.hubs.ly/H0j-Qpq0

We have an event every week these days! 🍕 After the machine learning challenge, we organized a sprint with Unaite to onboard students to open source. Open source is a great way to showcase their skills, as would a personal project!