Mailman 3 March 2022 - scikit-learn

[ANN] scikit-learn 0.24.0rc1 is online!
by Olivier Grisel Jan. 19, 2024

Jan. 19, 2024

Please help us test the first release candidate for scikit-learn 0.24.0: pip install scikit-learn==0.24.0rc1 Changelog: https://scikit-learn.org/0.24/whats_new/v0.24.html In particular, if you maintain a project with a dependency on scikit-learn, please let us know about any regression. Feel free to also retweet the announcement to get more people to test it before the final release (probably in 1 week or 2): https://twitter.com/scikit_learn/status/1334562221498753026 Thanks to anybody who helped make this happen! -- Olivier

5 12

[ANNOUNCEMENT] scikit-learn 1.0 release
by Adrin June 30, 2023

June 30, 2023

Hi everyone, We're happy to announce the 1.0 release which you can install via pip or conda: pip install -U scikit-learn or conda install -c conda-forge scikit-learn You can read the release highlights under https://scikit-learn.org/stable/auto_examples/release_highlights/plot_relea… and the long list of the changes under https://scikit-learn.org/stable/whats_new/v1.0.html New major features include: mandatory keyword arguments in many places, Spline Transformers, Quantile Regressor, Feature Names Support, a more flexible plotting API, Online One-Class SVM, and much more! This version supports Python versions 3.7 to 3.9. A big thanks to all contributors for making this release possible. Regards, Adrin, On the behalf of the scikit-learn maintainer team.

13 19

scikit-learn monthly developer meeting: Monday March 28 2022
by Thomas J. Fan April 1, 2022

April 1, 2022

Dear all, The scikit-learn developer monthly meeting will take place on Monday March 28 at 20:00 UTC. - Video call link: https://meet.google.com/ews-uszu-djs - Meeting notes / agenda: https://hackmd.io/0yokz72CTZSny8y3Re648Q?both - Local times: https://www.timeanddate.com/worldclock/meetingdetails.html?year=2022&month=… The goal of this meeting is to discuss ongoing development topics for the project. Everybody is welcome. As usual, please follow the code of conduct of the project: https://github.com/scikit-learn/scikit-learn/blob/main/CODE_OF_CONDUCT.md Regards, Thomas

2 1

Nystroem kernel approximation on precomputed kernel matrix
by Kab Kalaichelvan March 24, 2022

March 24, 2022

Dear scikit-learn My name is Kabiltan Kalaichelvan and I am an undergraduate researcher working under Professor Dane Morgan at the University of Wisconsin Madison. I am wondering if there are any code examples of Nystroem reducing the dimensionality of a precomputed kernel matrix(i.e. passing in 'precomputed' for the kernel argument). Thank you for your time. Best regards, Kabiltan

1 0

scikit-learn: March 2022 Updates: blog(s) + more
by Reshama Shaikh March 13, 2022

March 13, 2022

Hello, Read on for March 2022 updates from scikit-learn. [1] scikit-learn has a * BLOG * ! [2] Blog: Performance in scikit-learn [3] Blog: Three Components for Reviewing a Pull Request [4] Calendar: Office Hours and Meetings [5] scikit-learn is on social media: support us & get updates by following us! [6] scikit-learn MOOC is ongoing and you can still sign-up! [7] NumFOCUS Open Source Research Project [1] We would like to share a significant milestone in scikit-learn: a BLOG! https://blog.scikit-learn.org This blog was created using open source tools such as Jekyll and GitHub Pages. Lauren Burke, the blog creator, will be sharing how to get started creating a blog and hosting it for free. Sign-up is here for the March 22 event: https://www.meetup.com/data-umbrella/events/284042132/ >Jekyll is a static site generator that can be used to create a custom website simply, efficiently, and for free of charge. In this session, you will learn to set up a Jekyll-based website and blog, install a basic theme, add customizations, and host it via GitHub Pages. [2] Maintainer Julien Jerphanion has written a few articles on performance in scikit-learn. >scikit-learn has been around for more than 10 years. Yet, scikit-learn has some room to maneuver when it comes to performance. This series of blog posts aim to explain the on-going work of the scikit-learn developers to boost the library’s performance. https://blog.scikit-learn.org/technical/performances/ [3] Maintainer Thomas Fan has presented on etiquette in open source (video + slides + blog), "Three Components for Reviewing a Pull Request": 1. The mechanics of code review on GitHub. 2. The social aspects of code review and how to effectively give feedback. 3. The technical aspects of reviewing a pull request. Blog: https://blog.scikit-learn.org/community/pull-request/ Video: https://youtu.be/dyxS9KKCNzA [4] There is a handy calendar available on the blog website with meetings (Community Office Hours, Triage Team Meeting, Monthly Developer Meetings). https://blog.scikit-learn.org/calendar/ [5] Connect with scikit-learn on Social Media by following us on your favorite platforms! Feel free to share with others in your community including colleagues and students. [a] Twitter: https://twitter.com/scikit_learn [b] Twitter (commits): https://twitter.com/sklearn_commits [c] LinkedIn: https://www.linkedin.com/company/scikit-learn/ [d] YouTube: https://www.youtube.com/channel/UCJosFjYm0ZYVUARxuOZqnnw/playlists [e] Facebook: https://www.facebook.com/scikitlearnofficial/ [f] Instagram: https://www.instagram.com/scikitlearnofficial/ (photos worth seeing here!) [g] TikTok: https://www.tiktok.com/@scikit.learn [6] MOOC The second edition of the scikit-learn MOOC, Machine Learning in Python with scikit-learn, is open (registration <https://www.fun-mooc.fr/en/courses/machine-learning-python-scikit-learn/>). This *free* online course runs from February 15 to May 17, 2022. It is beginner-friendly, and a strong technical background is not required. Learners should have some familiarity with numpy, pandas, matplotlib. A certificate will be issued upon completion. There is still time to join this course. [7] NumFOCUS Open Source Research Project scikit-learn is working with NumFOCUS <http://numfocus.org> on a research project <https://numfocus.org/diversity-inclusion-disc/a-pivotal-time-in-numfocuss-p…> funded by the Gordon & Betty Moore Foundation <https://www.moore.org/> to understand the barriers to participation that contributors, particularly those from historically underrepresented groups, face in the open-source software community. The research team would like to talk to new contributors, project developers and maintainers, and those who have contributed in the past about their experiences joining and contributing to scikit-learn. Interested in sharing your experiences? Please complete this brief “Participant Interest <https://numfocus.typeform.com/to/WBWVJSqe>” form which contains additional information on the research goals, privacy, and confidentiality considerations. Your participation will be valuable to the growth and sustainability of diverse and inclusive open-source software communities. Accepted participants will participate in a 30-minute interview with a research team member. Cheers, Reshama & Lauren https://scikit-learn.org/dev/about.html

1 0

SGD Early Stopping
by Robert Slater March 9, 2022

March 9, 2022

We have something we are not understanding. clf2 = SGDClassifier(loss='log', penalty='l2',shuffle=True, max_iter=10,tol=.00001, early_stopping=True, validation_fraction=0.2, n_iter_no_change=2, verbose=0, random_state=1) clf2.fit(X_train,y_train) clf2.n_iter_ The result of the last line is ALWAYS n_iter_no_chang+1. (in this case 3, if we set n_iter+no_change=10, it ends at 11) No matter how I try to slow things down, it appears the early stopping kicks in at epoch 1. We've played with the learning rate, tolerance, etc... to try and make sure our problem isn't being solved in 1 epoch (which does seem dubious). I even ran this manually and scored the accuracy (along with enabling warm_start=True, and max_iter=1) for i in range(5): clf2.fit(X_train,y_train) p = clf2.predict(X_test) print(accuracy_score(p,y_test)) 0.9748226138704509 0.987182421606775 0.9881742580300603 0.9879453727016099 0.991760128175784 So it seems there is some accuracy improvement there to be had, however small--We're stumped as to what is going on and could use some wiser minds to explain this behavior

1 0

Does KernelDensity works for a vector with two elements?
by Mahmood Naderan March 1, 2022

March 1, 2022

Hi, I would like to use KernelDensity for some vectors. If the length of vectors is greater than 2, there is no problem. However, for the following example, it seems that the density estimation doesn't work properly. v = [2.46415e+07,1.23208e+07] a = array(v).reshape(-1, 1) kde = KernelDensity(kernel='gaussian', bandwidth=1).fit(a) s = linspace(min(a),max(a)) e = kde.score_samples(s.reshape(-1,1)) plot(s, e) mi = argrelextrema(e, np.less)[0] print ("Minima:", s[mi]) The s[mi] is empty in the end. But indeed the plot shows a minima because there is a gap between two numbers. Is there any restriction or note about using KernelDensity? Regards, Mahmood

1 0