From mahmood.nt at gmail.com Tue Mar 1 06:34:28 2022 From: mahmood.nt at gmail.com (Mahmood Naderan) Date: Tue, 1 Mar 2022 12:34:28 +0100 Subject: [scikit-learn] Does KernelDensity works for a vector with two elements? Message-ID: 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 From rdslater at gmail.com Tue Mar 8 20:02:21 2022 From: rdslater at gmail.com (Robert Slater) Date: Tue, 8 Mar 2022 19:02:21 -0600 Subject: [scikit-learn] SGD Early Stopping Message-ID: 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 -------------- next part -------------- An HTML attachment was scrubbed... URL: From reshama.stat at gmail.com Sun Mar 13 09:15:45 2022 From: reshama.stat at gmail.com (Reshama Shaikh) Date: Sun, 13 Mar 2022 09:15:45 -0400 Subject: [scikit-learn] scikit-learn: March 2022 Updates: blog(s) + more Message-ID: 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 ). 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 on a research project funded by the Gordon & Betty Moore Foundation 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 ? 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 -------------- next part -------------- An HTML attachment was scrubbed... URL: From thomasjpfan at gmail.com Wed Mar 23 18:53:33 2022 From: thomasjpfan at gmail.com (Thomas J. Fan) Date: Wed, 23 Mar 2022 18:53:33 -0400 Subject: [scikit-learn] scikit-learn monthly developer meeting: Monday March 28 2022 Message-ID: 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=3&day=28&hour=20&min=0&sec=0&p1=1440&p2=240&p3=248&p4=195&p5=179&p6=224 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 -------------- next part -------------- An HTML attachment was scrubbed... URL: From kalaichelvan at wisc.edu Wed Mar 23 23:51:46 2022 From: kalaichelvan at wisc.edu (Kab Kalaichelvan) Date: Thu, 24 Mar 2022 03:51:46 +0000 Subject: [scikit-learn] Nystroem kernel approximation on precomputed kernel matrix Message-ID: 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 -------------- next part -------------- An HTML attachment was scrubbed... URL: