Surprise changelog

Changelog History

Page 1

• v1.1.3 Changes

  Date: 23/09/2022

  🐛 Bug Fixes

  • 🛠 Fixed a serialization bug (introduced in 1.1.2) that would make dump() fail.

• v1.1.2 Changes

  Date: 29/08/2022 This version requires Python >= 3.8.

  ✨ Enhancements

  • 🐎 Significantly improve performance of SVD (~5X) and SVDpp (~10X) - by Hercules Smith
  • Minor improvements to similarity computations
  • Faster nearest neighbor search

  🐛 Bug Fixes

  • ✅ Fixed 'rank_test_fcp' order in grid search's cv_results
  • 🖐 Handle 0 ratings and avoid ZeroDivisionError in NMF
  • 🛠 (Hopefully) fixes various numpy and Cython compatibility issues with different Python versions (particularly 3.10)

• v1.1.1 Changes

  Date: 19/07/2020

  ✏️ Mostly doc typos and some minor bug fixes. Future versions (if any) 👍 will not support Python 2 anymore.

  🐛 Bug Fixes

  • 'mse' is now available in GridSearCV and RandomizedSearchCV
  • ⚡️ The Jester dataset link was updated
  • 🛠 Fixed a potential race condition when creating dataset directories

• v1.1.0 Changes

  September 13, 2019

  Date: 13/11/2019

  1.1.0 will be the last stable version with new features. Next versions will 🛠 only provide bug-fixes, but no new features. (And probably not support Python 2 at all).

  ✨ Enhancements

  • The prompt confirmation can now be disabled when downloading a dataset.
  • The MSE metric has been added.

  🐛 Bug Fixes

  • 🛠 Fixed a bug where msd and peasron would not properly set the similarity to zero when min_support wasn't reached.

  API Changes

  • 🗄 Tools that were deprecated before (data.split(), GridSearch, evaluate) are now removed.

• v1.1.rc0

  September 13, 2019

• v1.0.6 Changes

  April 22, 2018

  Date: 22/04/18

  ✨ Enhancements

  • ➕ Added new RandomizedSearchCV by David Stevens
  • ➕ Added verbose option to algorithms using a similarity matrix or baseline computation, to avoid unwanted printed messages.
  • When PredictionImpossible is raised, the prediction is now deferred to default_prediction() method, which can be overridden is child classes. This allows to not always set the default prediction to the average rating, which can be useful for some algorithms (e.g. those working with implicit positive feedback).
  • LeaveOneOut() now accepts a min_n_ratings parameter to make sure users in the trainset have at least min_n_ratings ratings.
  • Dumping is now done with pickle's highest protocol which allows for larger files.

  🐛 Bug Fixes

  • 0️⃣ Joblib parameter n_jobs now defaults to 1 (no use of multiprocessing). Should fix issues with Windows users.
  • cross_validate now returns correct values for training measures (used to return test measures instead).

• v1.0.5 Changes

  January 09, 2018

  Date: 09/01/18

  ✨ Enhancements

  • Cross-validation tools have been entirely reworked. We can now rely on powerful and flexible cross-validation iterators, inspired by scikit-learn's API.
  • the evaluate() method has been replaced by cross-validate which is parallel and can return measures on trainset as well as computation times.
  • 👷 GridSearch is now parallel, using joblib.
  • GridSearch now allows to refit an algorithm on the whole dataset.
  • 0️⃣ default data directory can now be custom with env variable SURPRISE_DATA_FOLDER
  • the fit() (and train()) methods now return self, which allows one-liners like algo.fit(trainset).test(testset)
  • Algorithms using a random initialization (e.g. SVD, NMF, CoClustering) now have a random_state parameter for seeding the RNG.
  • The getting started guide has been rewritten

  API Changes

  • 🗄 The train() method is now deprecated and replaced by the fit() method (same signature). Calls to train() should still work as before.
  • 🗄 Using data.split() or accessing the data.folds() generator is deprecated and replaced by the use of the more powefull CV iterators.
  • evaluate() is deprecated and replaced by model_selection.cross_validate(), which is parallel.
  • 🗄 GridSearch is deprecated and replaced by model_selection.GridSearchCV()

• v1.0.4 Changes

  September 30, 2017

  Date: 20/09/17

  ✨ Enhancements

  • ➕ Added possibility to load a dataset from a pandas dataframe
  • ➕ Added Precision and Recall examples to the FAQ (Maher Malaeb)
  • ➕ Added a kNN algorithm with normalization by z-score (Hengji Liu)
  • kNN algorithms now use heapq instead of list.sort() (computation time enhancement for large datasets).

  🛠 Fixes

  • Prediciont.str() when r_ui is None
  • GridSearch for dict parameters is now working as expected

  API Changes

  • 👀 param_grid for GridSearch is now slightly different for dict parameters (see note on the docs).

• v1.0.3 Changes

  May 03, 2017

  Date: 03/05/17

  ✨ Enhancements

  • ➕ Added FAQ in the doc
  • ➕ Added the possibility to retrieve the k nearest neighbors of a user or an item.
  • 🔄 Changed the dumping process a bit (see API changes). Plus, dumps can now be loaded.
  • ➕ Added possibility to build a testset from the ratings of a training set
  • ➕ Added inner-to-raw id conversion in the Trainset class
  • The r_ui parameter of the predict() method is now optional

  🛠 Fixes

  • 🛠 Fixed verbosity of the evaluate function
  • Corrected prediction when only user (or only item) is unknown in SVD and NMF algorithms. Thanks to kenoung!
  • Corrected factor vectors initialization of SVD algorithms. Thanks to adideshp!

  API Changes

  • The dump() method now dumps a list of predition (optional) and an algorithm (optional as well). The algorithm is now a real algorithm object. The trainset is not dumped anymore as it is already part of the algorithm anyway.
  • The dump() method is now part of the dump namespace, and not the global namespace (so it is accessed by surprise.dump.dump)

• v1.0.2 Changes

  January 04, 2017

  Date: 04/01/17

  👍 Just a minor change so that README.md is converted to rst for better rendering on PyPI.

• 1
• 2
• Next ›
• Last »