shrink

Prune regression ensemble

Syntax

cmp = shrink(ens)

cmp = shrink(ens,Name=Value)

Description

cmp = shrink(ens) returns a CompactRegressionEnsemble version of ens, a regularized regression ensemble model. cmp retains only those learners with weights above the specified threshold. shrink orders the members of cmp from largest to smallest. You can make predictions using cmp in the same way as when using ens with the predict function.

cmp = shrink(ens,Name=Value) specifies additional options using one or more name-value arguments. For example, you can specify the regularization parameter values for lasso, a lower cutoff on weights for weak learners, and the column index containing learner weights.

Examples

Shrink Bagged Regression Ensemble

Shrink a 300-member bagged regression ensemble, and view the number of members of the resulting ensemble.

Generate sample data.

X = rand(2000,20);
Y = repmat(-1,2000,1);
Y(sum(X(:,1:5),2)>2.5) = 1;

Shrink a 300-member bagged regression ensemble using 0.1 for the parameter Lambda.

bag = fitrensemble(X,Y,Method="Bag",NumLearningCycles=300);
cmp = shrink(bag,Lambda=0.1);

View the number of members of the resulting ensemble.

cmp.NumTrained

ans = 
94

Input Arguments

ens — Regression ensemble model

RegressionEnsemble model object | RegressionBaggedEnsemble model object

Regression ensemble model, specified as a RegressionEnsemble or RegressionBaggedEnsemble model object trained with fitrensemble.

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: shrink(ens,Threshold=0.1,Weightcolumn=1) specifies to use a lower cutoff of 0.1 for weak learners, and that the first column of ens contains learner weights.

Lambda — Regularization parameter value

"[]" (default) | nonnegative scalar

Regularization parameter value for lasso, specified as a nonnegative scalar. If ens contains a nonempty Regularization structure (that is, the property contains values specified by the regularize function), you cannot specify lambda. In this case, shrink regularizes ens using ens.Regularization.Lambda.

Example: Lambda=0.1

Data Types: single | double

Threshold — Lower cutoff on weights for weak learners

0 (default) | numeric nonnegative scalar

Lower cutoff on weights for weak learners, specified as a numeric nonnegative scalar. shrink creates cmp from those learners with weights above Threshold.

Example: Threshold=0.1

Data Types: single | double

Weightcolumn — Weights column

1 (default) | positive integer

Weights column, specified as a positive integer. Weightcolumn is the column index of ens.Regularization.TrainedWeights. shrink creates cmp with learner weights from this column.

Example: Weightcolumn=2

Data Types: single | double

Extended Capabilities

GPU Arrays

Accelerate code by running on a graphics processing unit (GPU) using Parallel Computing Toolbox™.

This function fully supports GPU arrays. For more information, see Run MATLAB Functions on a GPU (Parallel Computing Toolbox).

Version History

Introduced in R2011a