1.3 KiB
1.3 KiB
Bootstrap Aggregator
Bootstrap Aggregating (or bagging for short) is a model averaging technique designed to improve the stability and performance of a user-specified base estimator by training a number of them on a unique bootstrapped training set sampled at random with replacement. Bagging works especially well with estimators that tend to have high prediction variance by reducing the variance through averaging.
Interfaces: Estimator, Learner, Parallel, Persistable
Data Type Compatibility: Depends on base learner
Parameters
| # | Name | Default | Type | Description |
|---|---|---|---|---|
| 1 | base | Learner | The base learner. | |
| 2 | estimators | 10 | int | The number of base learners to train in the ensemble. |
| 3 | ratio | 0.5 | float | The ratio of samples from the training set to randomly subsample to train each base learner. |
Example
use Rubix\ML\BootstrapAggregator;
use Rubix\ML\Regressors\RegressionTree;
$estimator = new BootstrapAggregator(new RegressionTree(10), 300, 0.2);
Additional Methods
This meta estimator does not have any additional methods.