0.19.0 - 2023-08-02

Calling learn_one in a pipeline will now update each part of the pipeline in turn. Before the unsupervised parts of the pipeline were updated during predict_one. This is more intuitive for new users. The old behavior, which yields better results, can be restored by calling learn_one with the new compose.learn_during_predict context manager.

bandit

• Added a bandit.datasets submodule, which is meant to contain contextual bandit datasets.
• Added bandit.base.ContextualPolicy.
• Added bandit.datasets.NewsArticles.
• Added bandit.LinUCBDisjoint, which is River's first contextual bandit policy.
• Added bandit.RandomPolicy.

compose

• Removed the compose.warm_up_mode context manager.
• Removed the compose.pure_inference_mode context manager.
• The last step of a pipeline will be correctly updated if it is unsupervised, which wasn't the case before.
• Fixed an edge-case where compose.TransformerProduct would not work when chained more than twice.

drift

• Added a datasets submodule, which contains datasets that are useful for concept drift experiments.
• Fix bugs in drift.binary.HDDM_A and drift.binary.HDDM_W.

linear_model

• Added a predict_many method to linear_model.BayesianLinearRegression.
• Added a smoothing parameter to linear_model.BayesianLinearRegression, which allows it to cope with concept drift.

forest

• Fixed issue with forest.ARFClassifier which couldn't be passed a CrossEntropy metric.
• Fixed a bug in forest.AMFClassifier which slightly improves predictive accuracy.
• Added forest.AMFRegressor.

multioutput

• Added metrics.multioutput.SampleAverage, which is equivalent to using average='samples' in scikit-learn.

preprocessing

• Added preprocessing.OrdinalEncoder, to map string features to integers.
• The transform_many method of preprocessing.StandardScaler now uses the dtype of the input for the output.

proba

• Added proba.MultivariateGaussian.

stream

• stream.iter_arff now supports sparse data.
• stream.iter_arff now supports multi-output targets.
• stream.iter_arff now supports missing values indicated with question marks.

utils

• Added utils.random.exponential to retrieve random samples following an exponential distribution.