learnersize

Compact size of trained machine learning model object

Since R2024b

Syntax

size = learnersize(Mdl)

size = learnersize(Mdl,SizeType=type)

Description

size = learnersize(Mdl) returns the compact size (in bytes) of the trained machine learning model object Mdl. The learnersize function calculates the model size by calling gather and whos. If Mdl is compact or has no supported compact object version, learnersize returns its actual size. Otherwise, learnersize returns the size of the compact version of Mdl.

size = learnersize(Mdl,SizeType=type) returns the model size of the specified type. If you specify type="coder" and Mdl is supported for code generation, learnersize returns the approximate size of Mdl (in bytes) in C/C++ code generated by saveLearnerForCoder and codegen (MATLAB Coder).

Examples

Determine Compact Model Size

Train a neural network classification model using the ionosphere data set, and return the compact model object size in bytes.

load ionosphere.mat;
Mdl = fitcnet(X,Y);
size = learnersize(Mdl)

size = 
8142

Determine Approximate Code-Generated Model Size

Train a neural network classification model using the ionosphere data set.

load ionosphere.mat;
Mdl = fitcnet(X,Y);

Return the approximate size of the model (in bytes) in C/C++ code generated by saveLearnerForCoder and codegen (MATLAB Coder).

size = learnersize(Mdl,SizeType="coder")

size = 
3834

Because the learnersize function does not generate code, the size represents an estimate of the amount of system memory that would be used by the trained model in generated code.

Input Arguments

Mdl — Machine learning model

regression model object | classification model object

Machine learning model, specified as a regression or classification model object as given in the following tables of supported models. If you specify SizeType="coder", then Mdl must be supported for code generation.

Classification Model Object

Model	Full and Compact Model Objects	Code Generation Support
Discriminant analysis classification	ClassificationDiscriminant CompactClassificationDiscriminant	Yes
Multiclass model for support vector machines (SVMs) or other classifiers	ClassificationECOC CompactClassificationECOC	Yes
Ensemble classifier	ClassificationEnsemble CompactClassificationEnsemble ClassificationBaggedEnsemble	Yes
Gaussian kernel model for binary classification of big data	ClassificationKernel	Yes
k-nearest neighbor classification	ClassificationKNN	Yes
Linear model for binary classification of high-dimensional data	ClassificationLinear	Yes
Naive Bayes classifier	ClassificationNaiveBayes CompactClassificationNaiveBayes	Yes
Neural network classifier	ClassificationNeuralNetwork CompactClassificationNeuralNetwork	Yes
SVM for one-class and binary classification	ClassificationSVM CompactClassificationSVM	Yes
Binary decision tree for classification	ClassificationTree CompactClassificationTree	Yes
Generalized additive model	ClassificationGAM CompactClassificationGAM	No

Regression Model Object

Model	Full and Compact Model Object	Code Generation Support
Generalized linear model	GeneralizedLinearModel CompactGeneralizedLinearModel	Yes
Linear regression model	LinearModel CompactLinearModel	Yes
Ensemble regression	RegressionEnsemble CompactRegressionEnsemble RegressionBaggedEnsemble	Yes
Gaussian process regression	RegressionGP CompactRegressionGP	Yes
Gaussian kernel regression for big data	RegressionKernel	Yes
Linear regression for high-dimensional data	RegressionLinear	Yes
Neural network regression	RegressionNeuralNetwork CompactRegressionNeuralNetwork	Yes
Regression tree	RegressionTree CompactRegressionTree	Yes
SVM regression	RegressionSVM CompactRegressionSVM	Yes
Generalized additive model	RegressionGAM CompactRegressionGAM	No

type — Type of size output

"compact" (default) | "coder"

Type of the size output, specified as one of the values in this table:

Value	Output of learnersize
"compact" (default)	Compact size (in bytes) of the trained machine learning model object Mdl. If Mdl is compact or has no supported compact object version, learnersize returns its actual size. Otherwise, learnersize returns the size of the compact version of Mdl.
"coder"	Approximate size of Mdl (in bytes) in C/C++ code generated by saveLearnerForCoder and codegen. Because the learnersize function does not generate code, the size represents an estimate of the amount of system memory that would be used by the trained model in generated code.

If you specify SizeType="coder", then Mdl must be supported for code generation (see saveLearnerForCoder).

Data Types: char | string

Version History

Introduced in R2024b