Mailman 3 November 2017 - scikit-learn

Topic for thesis work on scikit learn
by Gaurav Dhingra Dec. 31, 2017

Dec. 31, 2017

Hi everyone, I am a final year (5th year) undergraduate Applied Mathematics student in India. I am thinking of doing my final year thesis by doing some work (coding part) on scikit learn, so I was thinking if anyone could tell me if there are available topics (not necessarily names of those topics) that I could work on being an undergraduate student? I would want to expand upon this in December when my exams will be over. But in the mean time would want to take a step in that direction by just knowing if there will be available topics that I could work on. It could be the case that available topics are not so easy for an undergraduate, still in that case I would like to do some research on the topics first. -- Best, Gaurav Dhingra (sent from Thunderbird email client)

2 4

Webservice that uses scikit-learn
by Orges Leka Nov. 30, 2017

Nov. 30, 2017

Dear scikit-learn developers, I have developed a small webservice which can hold multiple scikit-learn models and serve post - json requests for prediction. A model must have model.metadata and must implement model.transform_predict(newdata). There are two examples: BostonModel, where only predict is overriden from WebModel IrisModel, where predict and transform is overriden from WebModel. The idea is, that while fitting a model, you could have some metadata which are needed for prediction. These metadata are stored as a python dictionary. metadata could hold for example: version of model when it was created additional pandas.DataFrames needed for prediction constants needed in the predict computation metrics about the model etc. The repo can be found here: https://github.com/orgesleka/webscikit It comes with two examples: iris and boston. The server can load other models at runtime, in case one is changing the models. The repo is meant as a proof of concept. If somebody has ideas on how to improve things or adding new features, that would be great. To get started, see: https://github.com/orgesleka/webscikit/wiki/Getting-started Kind regards Orges Leka

1 0

Re: [scikit-learn] 1. Re: Rapid Outlier Detection via Sampling (Olivier Grisel)
by Orges Leka Nov. 28, 2017

Nov. 28, 2017

Dear Olivier and Gael , Thank you for your answer. I started a request for inclusion in scikit-learn-contrib. The repo can be found here: https://github.com/orgesleka/rapid-outlier-detection Kind regards Orges Leka 2017-11-27 18:00 GMT+01:00 <scikit-learn-request(a)python.org>: > Send scikit-learn mailing list submissions to > scikit-learn(a)python.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://mail.python.org/mailman/listinfo/scikit-learn > or, via email, send a message with subject or body 'help' to > scikit-learn-request(a)python.org > > You can reach the person managing the list at > scikit-learn-owner(a)python.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of scikit-learn digest..." > > > Today's Topics: > > 1. Re: Rapid Outlier Detection via Sampling (Olivier Grisel) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Mon, 27 Nov 2017 09:45:22 +0100 > From: Olivier Grisel <olivier.grisel(a)ensta.org> > To: Scikit-learn mailing list <scikit-learn(a)python.org> > Subject: Re: [scikit-learn] Rapid Outlier Detection via Sampling > Message-ID: > <CAFvE7K7xzpqfJgCU6OCZ8octnF9N0XUM9CDQ0+mDfmLbyKWrBw@mail. > gmail.com> > Content-Type: text/plain; charset="utf-8" > > > Do I need to write object oriented or are functions also ok? > > I you want to contribute an implementation as a new project on scikit-learn > contrib, you should be careful to follow the scikit-learn estimators API: > > http://scikit-learn.org/dev/developers/contributing.html# > apis-of-scikit-learn-objects > > For outlier detection in particular, you should make sure your new > estimator is consistent with the API conventions of other methods already > in scikit-learn: > > http://scikit-learn.org/dev/modules/outlier_detection.html > > One of the primary goals of the scikit-learn ecosystem is to provide a > simple homogeneous API to a very heterogeneous set of methods. > > -- > Olivier > ? >

1 0

Jeff Levesque: sklearn + D3JS
by Jeffrey Levesque Nov. 27, 2017

Nov. 27, 2017

Hi, I'm developing an API for sklearn: - https://github.com/jeff1evesque/machine-learning I was wondering if anyone had integrated visualization tools, like D3JS, with results from sklearn predictions? If so, would any of you be willing to show how the backend results was piped into JavaScript? PS. If anyone is willing to contribute, or help in anyway, the codebase is BSD. Thank you, Jeff Levesque https://github.com/jeff1evesque

1 0

Rapid Outlier Detection via Sampling
by Orges Leka Nov. 27, 2017

Nov. 27, 2017

Dear scikit-learn Developers, My Name is Orges Leka and I would like to implement "Rapid Outlier Detection via Sampling" [1] in scikit-learn. In R this method is already available [2] by the authors of the method. In Python I have not seen any implementation yet. The method is very simple yet effective as the authors show. First one selects say 20 points. Then computes the shortest distance of all other points to these 20 points. This is the outlier-score for one specific point. It would be nice to implement this with different metrics / distances (euclid, manhattan or other metrics) . How would I start the implementation? I have already git-cloned scikit-learn on my pc. Do I need to write object oriented or are functions also ok? If this succeeds, I would also like to extend the "example-outliers" doc with the above method. Kind regards Dipl. Math. Orges Leka [1] https://papers.nips.cc/paper/5127-rapid-distance-based-outlier-detection-vi… [2] https://github.com/mahito-sugiyama/sampling-outlier-detection

3 2

Issue with Sihouette_samples
by Shiheng Duan Nov. 16, 2017

Nov. 16, 2017

Hi all, I am doing cluster work and wanna use silhouette score to determine the number of clusters. But I got MemoryError when execute silhouette_samples. I searched it and found something related to numpy. But I cannot reproduce the numpy error. Is there any solution to it? The data is 621*1405*12. Thanks!

3 3

Custom Distance Metric / Distance Matrix with K-means?
by Shane Grigsby Nov. 15, 2017

Nov. 15, 2017

Hello, I'd like to be able to cluster data using either k-means or mini-batch-kmeans for a toroidal geometry. I know that if I was using DBSCAN I could pass in a pre-computed distance matrix to do this; if I was using OPTICS I could pass in a 'metric' keyword for distance and specify a custom distance metric. Is this possible for K-means / minibatch-kmeans? I don't see distance metrics documented as possible keyword arguments... but perhaps they're allowed as **kwargs that pass to the underlying distance calculation call? Thanks, Shane -- *PhD candidate & Research Assistant* *Cooperative Institute for Research in Environmental Sciences (CIRES)* *University of Colorado at Boulder*

3 3

New core devs: Hanmin Qin, Guillaume Lemaître, and Roman Yurchak
by Gael Varoquaux Nov. 10, 2017

Nov. 10, 2017

Hi scikit-learn community, A week ago, we added 3 core developers, but I think that we forgot to announce it. So let me please welcome on board Hanmin Qin, Guillaume Lemaître, and Roman Yurchak. They have been very active in the development of the project, and very helpful in the review process. It's a pleasure to see the team growing. Gaël

3 2

Question about dummy coding using DictVectorizer or FeatureHasher: generating correlated dimensions
by Yusuke Nishioka Nov. 7, 2017

Nov. 7, 2017

Hello, I have a question about dummy coding using DictVectorizer or FeatureHasher. ``` >>> from sklearn.feature_extraction import DictVectorizer, FeatureHasher >>> D = [{'age': 23, 'gender': 'm'},{'age': 34, 'gender': 'f'},{'age': 18, 'gender': 'f'},{'age': 50, 'gender': 'm'}] >>> m1 = FeatureHasher(n_features=10) >>> m1.fit_transform(D).toarray() array([[ 0., 0., -1., 0., 0., 0., 0., 0., 0., 23.], [ 0., 0., 0., 0., 0., 0., 0., 0., 1., 34.], [ 0., 0., 0., 0., 0., 0., 0., 0., 1., 18.], [ 0., 0., -1., 0., 0., 0., 0., 0., 0., 50.]]) >>> m2 = DictVectorizer(sparse=False) >>> m2.fit_transform(D) array([[ 23., 0., 1.], [ 34., 1., 0.], [ 18., 1., 0.], [ 50., 0., 1.]]) >>> m2.feature_names_ ['age', 'gender=f', 'gender=m'] ``` Since both DictVectorizer and FeatureHasher generate dimensions for 'gender=m' and 'gender=f', these dimensions are perfectly correlated. This is because DictVectorizer and FeatureHasher by default generate n dimensions for n categorical values of 1 feature. My questions are as follows: 1. My expectation is for them to generate n-1 dimensions for n categorical values, and is there any way to do this using DictVectorizer and FeatureHasher? 2. How should I handle these correlated dimensions? In my understanding, the training on data which has colinearity will make prediction unstable. Will L1 or L2 regularization work for this problem? If there is any issue or article related to these questions, would you please tell me the URL? Thank you. Regards, Yusuke

1 0