Skip to content


Successive halving algorithm for classification.

Successive halving is a method for performing model selection without having to train each model on all the dataset. At certain points in time (called "rungs"), the worst performing will be discarded and the best ones will keep competing between each other. The rung values are designed so that at most budget model updates will be performed in total.

If you have k combinations of hyperparameters and that your dataset contains n observations, then the maximal budget you can allocate is:


It is recommended that you check this beforehand. This bound can't be checked by the function because the size of the dataset is not known. In fact it is potentially infinite, in which case the algorithm will terminate once all the budget has been spent.

If you have a budget of B, and that your dataset contains n observations, then the number of hyperparameter combinations that will spend all the budget and go through all the data is:

\[\ceil(\floor(\frac{B}{(2n)}) \times eta)\]


  • models

    The models to compare.

  • metric (river.metrics.base.Metric)

    Metric used for comparing models with.

  • budget (int)

    Total number of model updates you wish to allocate.

  • eta – defaults to 2

    Rate of elimination. At every rung, math.ceil(k / eta) models are kept, where k is the number of models that have reached the rung. A higher eta value will focus on less models but will allocate more iterations to the best models.

  • verbose – defaults to False

    Whether to display progress or not.

  • print_kwargs

    Extra keyword arguments are passed to the print function. For instance, this allows providing a file argument, which indicates where to output progress.


  • best_model

    The current best model.


As an example, let's use successive halving to tune the optimizer of a logistic regression. We'll first define the model.

>>> from river import linear_model
>>> from river import preprocessing

>>> model = (
...     preprocessing.StandardScaler() |
...     linear_model.LogisticRegression()
... )

Let's now define a grid of parameters which we would like to compare. We'll try different optimizers with various learning rates.

>>> from river import utils
>>> from river import optim

>>> models = utils.expand_param_grid(model, {
...     'LogisticRegression': {
...         'optimizer': [
...             (optim.SGD, {'lr': [.1, .01, .005]}),
...             (optim.Adam, {'beta_1': [.01, .001], 'lr': [.1, .01, .001]}),
...             (optim.Adam, {'beta_1': [.1], 'lr': [.001]}),
...         ]
...     }
... })

We can check how many models we've created.

>>> len(models)

We can now pass these models to a SuccessiveHalvingClassifier. We also need to pick a metric to compare the models, and a budget which indicates how many iterations to run before picking the best model and discarding the rest.

>>> from river import expert

>>> sh = expert.SuccessiveHalvingClassifier(
...     models=models,
...     metric=metrics.Accuracy(),
...     budget=2000,
...     eta=2,
...     verbose=True
... )

A SuccessiveHalvingClassifier is also a classifier with a learn_one and a predict_proba_one method. We can therefore evaluate it like any other classifier with evaluate.progressive_val_score.

>>> from river import datasets
>>> from river import evaluate
>>> from river import metrics

>>> evaluate.progressive_val_score(
...     dataset=datasets.Phishing(),
...     model=sh,
...     metric=metrics.ROCAUC()
... )
[1] 5 removed       5 left  50 iterations   budget used: 500        budget left: 1500       best Accuracy: 80.00%
[2] 2 removed       3 left  100 iterations  budget used: 1000       budget left: 1000       best Accuracy: 84.00%
[3] 1 removed       2 left  166 iterations  budget used: 1498       budget left: 502        best Accuracy: 86.14%
[4] 1 removed       1 left  250 iterations  budget used: 1998       budget left: 2  best Accuracy: 84.80%
ROCAUC: 0.952889

We can now view the best model.

>>> sh.best_model
Pipeline (
  StandardScaler (
  LogisticRegression (
    optimizer=Adam (
      lr=Constant (
    loss=Log (
    intercept_lr=Constant (
    initializer=Zeros ()



Return a fresh estimator with the same parameters.

The clone has the same parameters but has not been updated with any data. This works by looking at the parameters from the class signature. Each parameter is either - recursively cloned if it's a River classes. - deep-copied via copy.deepcopy if not. If the calling object is stochastic (i.e. it accepts a seed parameter) and has not been seeded, then the clone will not be idempotent. Indeed, this method's purpose if simply to return a new instance with the same input parameters.


Update the model with a set of features x and a label y.


  • x (dict)
  • y (Union[bool, str, int])


Classifier: self


Predict the label of a set of features x.


  • x (dict)


typing.Union[bool, str, int]: The predicted label.


Predict the probability of each label for a dictionary of features x.


  • x


A dictionary that associates a probability which each label.