Regularization
For greater accuracy and link-function choices on low- through medium-dimensional data sets, fit a generalized linear model with a lasso penalty usinglassoglm
.
For reduced computation time on high-dimensional data sets, train a binary, linear classification model, such as a regularized logistic regression model, usingfitclinear
. You can also efficiently train a multiclass error-correcting output codes (ECOC) model composed of logistic regression models usingfitcecoc
.
For nonlinear classification with big data, train a binary, Gaussian kernel classification model with regularized logistic regression usingfitckernel
.
Classes
ClassificationLinear |
Linear model for binary classification of high-dimensional data |
ClassificationECOC |
Multiclass model for support vector machines (SVMs) and other classifiers |
ClassificationKernel |
Gaussian kernel classification model using random feature expansion |
ClassificationPartitionedLinear |
Cross-validated linear model for binary classification of high-dimensional data |
ClassificationPartitionedLinearECOC |
Cross-validated linear error-correcting output codes model for multiclass classification of high-dimensional data |
Functions
lassoglm |
Lasso or elastic net regularization for generalized linear models |
fitclinear |
Fit binary linear classifier to high-dimensional data |
templateLinear |
Linear classification learner template |
fitcecoc |
Fit multiclass models for support vector machines or other classifiers |
predict |
Predict labels for linear classification models |
fitckernel |
Fit binary Gaussian kernel classifier using random feature expansion |
predict |
Predict labels for Gaussian kernel classification model |
Examples and How To
- Regularize Poisson Regression
Identify and remove redundant predictors from a generalized linear model.
- Regularize Logistic Regression
Regularize binomial regression.
- Regularize Wide Data in Parallel
Regularize a model with many more predictors than observations.
Concepts
- Lasso Regularization of Generalized Linear Models
The lasso algorithm produces a smaller model with fewer predictors. The related elastic net algorithm can be more accurate when predictors are highly correlated.