Skip to content

Paris Sprint 25 Feb, 1 Mar 2019

Jump to bottom
Chiara Marmo edited this page Dec 9, 2019 · 2 revisions

We will be holding a sprint in Paris, 25 Feb to Mar 1st. Each day, the sprint will start at 9.30 am until 6.30 pm.

Contact: Guillaume Lemaitre - +33761104782

Location

AXA REV, 7th floor, Immeuble Java, 61 rue Mstislav Rostropovitch, 75017 Paris

Transport

  • Metro line 13 (Brochant / Porte de Clichy)
  • RER C (Porte de Clichy)
  • Tram T3b
  • Transilien line L (Pont Cardinet)

Social event

  • Monday 25 February: Breakfast at AXA with a short presentation (10 minutes) about sckit-learn and sprint objectives.
  • Friday 1 March: Aperitif at launch organised by AXA.

People present:

Indicate your expertise level, if you need funding for travel or accommodation and what you want to work on.

  • Gaël Varoquaux, core developer, no funding needed
  • Guillaume Lemaitre, core developer, no funding needed
  • Adrin Jalali, core developer, needs (some) funding
  • Alex Gramfort, core developer, no funding needed
  • Roman Yurchak, core developer, no funding needed
  • Joel Nothman (until 28 Feb), core developer, funding arranged, work_on={pandas issues, sample and feature props, review, too many things...}
  • Joan Massich, contributor, no funding needed
  • Joris Van den Bossche, core developer, no funding needed
  • Andreas Mueller, core developer, no funding needed, working on governance finalization, SLEP process, SLEP reviewing, Roadmap
  • Nicolas Hug, contributor, no funding needed
  • Thomas Fan, contributor, no funding needed
  • Nicolas Goix, contributor, no funding needed
  • Albert Thomas, contributor, no funding needed, might not be attending the whole week, merge PR #12827, refactor tests #10027, reviews
  • Jérémie du Boisberranger, contributor, no funding needed
  • Thomas Moreau, contributor, no funding needed
  • Pavel Soriano, new contributor, no funding needed
  • William de Vazelhes, contributor, no funding needed, PR #10058 in progress (metric learning)
  • Aurélien Bellet, contributor, no funding needed, will probably attend Mon-Tue, PR #8602 and #10058 (metric learning), issue #12228 (graph lasso)
  • Romuald Menuet, new contributor, no funding needed
  • Olivier Grisel, core developer, no funding need
  • Maria Telenczuk, new contributor, no funding needed.
  • Bartosz Telenczuk, new contributor, no funding needed.
  • Ivan Panico, contributor, no funding needed.
  • Oliver Rausch, contributor, funding has been handled.
  • Pierre Glaser, contributor, no funding needed.
  • Patricio Cerda, contributor, no funding needed (Online NMF).
  • Pierre Ablin, contributor, no funding needed (Online NMF).
  • Danilo Bzdok, no funding need (KNNImputer, AveragingRegressor, Added estimator checks for pandas object, FrequencyEncoder, Adding explained variances to sparse pca)
  • Sébastien Treguer, no funding needed. (joining if space available)
  • Assia Benbihi, new contributor, no funding needed.
  • Xavier Dupre, new contributor, no funding needed.
  • Samuel Ronsin, contributor, no funding needed.
  • Julien Jerphanion, potentail new contributor, no funding needed.

Suggested tasks

Welcoming new contributors

The sprint is a great time for new contributors to become familiar with the project. We welcome newcomers. Please be sure to read the contributing section of the documentation http://scikit-learn.org/dev/developers/contributing.html, and to have a development environment ready in which you can install scikit-learn from scratch, build it, and use git to push changes to github.

Technical Discussions Schedule

Time Monday Tuesday Wednesday Thursday Friday
10:00 Welcome (logreg) tol/convergence get_feature_names pandas handling efficient GridSearch
16:00 OPTICS? ARM? Freezing #9397 fit_transform sample props beers?

Discussions to add on in spare time:

  • GLM support... poisson regression, quantile regression, etc.
  • Euclidean distances consistency and stability
  • Sample props (and feature props?) and their transformation
  • contrib maintenance (is this the right model? who maintains it? what clear criteria for acceptance?)
  • keyword only arguments
  • pipeline slicing
  • search spaces

Explanation / issues:

  • Freezing:
  • convergence:
  • get_feature_names: (includes pipeline slicing because reasons)
  • fit_transform: (included imbalance learn interface discussion maybe)
  • search spaces: (related to configspace and searchgrid)
  • efficient grid-search: (should this include avoiding recomputation of preprocessing steps as well as the warm start logic? - our caching right now is not great and daskML does much better...

Meeting Minutes:

Clone this wiki locally