🌐 Mailman 3 September 2016 - scikit-learn - python.org

Large computation time for homogeneous data with agglomerative clustering
by Md. Khairullah Sept. 26, 2016

Sept. 26, 2016

Dear Scikit-learners, This is my first post here and I hope you experts can help me a lot. We are using the agglomerative clustering with ward's linkage and connectivity constraint. The data size is around 205,000 (each is a single scalar feature). The data set is dynamic (in time) and we need to apply clustering at different time thorough the process. Initially all data is 0 and they increase gradually. Alternatively, in the early stage the data is more homogeneous and the heterogeneity among the data increases gradually. If the clustering is applied at the final stage (most heterogeneous data, but off course having patterns/clusters) requesting 20 clusters it takes only 61s of CPU time. But, if clustering is run in an early stage (more homogeneous data but all are not 0 and off course there are patterns/clusters in the data) with the same settings the time rises up to 1h 5m. The CPU time is in-between of these two if the data come from an in-between time stamp. I also tried the the other linkage options too, but the situation does not improve. My understanding is that the homogeneity is playing the role. Have you experienced this too? What solution do you suggest? Thanks in advance for your attention and help. -- Best regards Md. Khairullah PhD Student, KU Leuven Numerical Analysis and Applied Mathematics Section Celestijnenlaan 200a - box 2402 3001 Leuven room: 03.18 tel. +32 16 37 39 66 fax +32 16 3 27996

1 0

Is there a built-in function for pairs of data?
by Afarin Famili Sept. 26, 2016

Sept. 26, 2016

Dear Scikit-learn team, We need to deal with pairs of data in our classification task. I was wondering if there is already a built-in function in Scikit-learn that can partition the pairs of data into train and test sets? Regards, Afarin ________________________________ UT Southwestern Medical Center The future of medicine, today.

3 2

header intact
by Afarin Famili Sept. 26, 2016

Sept. 26, 2016

? ________________________________ UT Southwestern Medical Center The future of medicine, today.

1 0

MLP Maxout activation proposal
by 深谷亮祐 Sept. 25, 2016

Sept. 25, 2016

Hi everyone, My name is Ryosuke Fukatani. I'm joining a scikit-learn community and really excited to work with all. Today I propose new feature "maxout activation" for MLP. Maxout activation achived high performance classification.(http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf). And it seemed that only a few layers are needed to use maxout. I think it is suitable feature for light weight MLP by scikit learn. If it's OK, I'd work on it. Best regards

2 4

Understanding the codebase
by Siddharth Gupta Sept. 24, 2016

Sept. 24, 2016

Hey fellows, I am new to Open Source, I have set up the dev-environment for scikit-learn but I dont know from where should I start understanding the codebase. I would be working on a continuous Integration issue, any support would be helpful. I have read the contribution guidelines. Regards Siddharth Gupta, Ph: 9871012292 Linkedin <https://www.linkedin.com/in/sidgupta234/> | Github <https://github.com/sidgupta234> | Codechef <https://www.codechef.com/users/sidgupta234> | Twitter <https://twitter.com/SidGupta234> | Facebook <https://www.facebook.com/profile.php?id=1483695876>

1 0

Call for reviewers for 0.18 release
by Andreas Mueller Sept. 22, 2016

Sept. 22, 2016

Hey folks. We are preparing to release 0.18 and we're a bit short on reviewers (I'm semi-online myself). If any of you want to pitch in, please have a look at the PRs with the 0.18 milestone: https://github.com/scikit-learn/scikit-learn/pulls?q=is%3Aopen+milestone%3A… Thanks, Andy

1 0

behaviour of OneHotEncoder somewhat confusing
by Lee Zamparo Sept. 22, 2016

Sept. 22, 2016

Hi sklearners, A lab-mate came to me with a problem about encoding DNA sequences using preprocessing.OneHotEncoder, and I find it to produce confusing results. Suppose I have a DNA string: myguide = ‘ACGT’ He’d like use OneHotEncoder to transform DNA strings, character by character, into a one hot encoded representation like this: [[1,0,0,0], [0,1,0,0], [0,0,1,0], [0,0,0,1]]. The use-case seems to be solved in pandas using the dubiously named get_dummies method ( http://pandas.pydata.org/pandas-docs/version/0.13.1/generated/pandas.get_du…). I thought that it would be trivial to do with OneHotEncoder, but it seems strangely difficult: In [23]: myarray = le.fit_transform([c for c in myguide]) In [24]: myarray Out[24]: array([0, 1, 2, 3]) In [27]: myarray = le.transform([[c for c in myguide],[c for c in myguide],[c for c in myguide]]) In [28]: myarray Out[28]: array([[0, 1, 2, 3], [0, 1, 2, 3], [0, 1, 2, 3]]) In [29]: ohe.fit_transform(myarray) Out[29]: array([[ 1., 1., 1., 1.], [ 1., 1., 1., 1.], [ 1., 1., 1., 1.]]) <— ???? So this is not at all what I expected. I read the documentation for OneHotEncoder ( http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneH…), but did not find if clear how it worked (also I found the example using integers confusing). Neither FeatureHasher nor DictVectorizer seem to be more appropriate for transforming strings into positional OneHot encoded arrays. Am I missing something, or is this operation not supported in sklearn? Thanks, -- Lee Zamparo

4 5

Contribution project proposal
by Iván Vallés Pérez Sept. 21, 2016

Sept. 21, 2016

Hello, My name is Iván Vallés and I am Data Scientist. I am really interested in contributing to the Scikit-Learn project by writing a new feature: Stacked Generalization Ensemble. I would like to know if it is being developed and if you think it is worth. Of course it is a long-term project due to the relative complexity of the proposal. Thank you. Best, Iván

4 4

Fwd: Featuring scikit-learn in Hacktoberfest 2016
by Andreas Mueller Sept. 16, 2016

Sept. 16, 2016

Hey all. Wdyt (see below)? I'm always for new contributors ;) - Though it does add overhead. I'm not sure I want to add a label, but being on the page seems a good thing, I think. Adny -------- Forwarded Message -------- Subject: Featuring scikit-learn in Hacktoberfest 2016 Date: Fri, 16 Sep 2016 17:31:30 -0400 From: Andrew Starr-Bochicchio <astarr(a)digitalocean.com> To: t3kcit(a)gmail.com CC: Open Source <opensource(a)digitalocean.com> Hi, My name is Andrew Starr-Bochicchio and I'm on the Community team at DigitalOcean. In an effort to create some excitement around contributing to open source in the month of October, DigitalOcean and GitHub are partnering to run the third annual Hacktoberfest. We challenge people to get involved by creating at least four pull requests to open source projects over the course of the month. Everyone who completes the task gets a limited edition t-shirt. Last year we saw 14,419 people sign up, and 5,708 people create at least four pull requests <https://www.digitalocean.com/company/blog/looking-back-at-hacktoberfest/>. We're hoping for even more this time around. Pull requests can be to any project, but we also like to use Hacktoberfest to feature a number of projects that we love, most prominently on the landing page (see last year's for an example <https://hacktoberfest.digitalocean.com/>). We are hoping to feature scikit-learn, but I just wanted to check in with you and make sure you were interested. Additionally, this year we'll be encouraging project maintainers to label issues on GitHub "hacktoberfest" so that we can get even more projects involved and send contributors to where the help is most needed. So spread the word if there are other projects you think might be interested. Please let me know if you'd like to be featured on the landing page, if you have any questions, or if there's anything DO can do to help. Thanks! - Andrew Starr-Bochicchio Community Manager DigitalOcean

3 4

New contributor to scikit-learn
by Kathleen Chen Sept. 16, 2016

Sept. 16, 2016

Hi! I'm Kathy, a student at Penn taking an open source software development <https://www.cis.upenn.edu/~cdmurphy/foss/fall2016/> course this semester. I wanted to send a quick "hello" to the community; a few other students and I will be making contributions to scikit-learn as part of our work in the class. :) Really looking forward to learning from & getting to know some of you in the coming months!

6 7