Lasso Regression

GPTKB entity

Statements (32)
Predicate Object
gptkbp:instanceOf Regression technique
gptkbp:alsoKnownAs gptkb:Least_Absolute_Shrinkage_and_Selection_Operator
gptkbp:assumes Linear relationship between variables
gptkbp:canSetCoefficientsToZero Yes
gptkbp:category gptkb:Supervised_learning
Linear model
gptkbp:contrastsWith gptkb:Elastic_Net
gptkb:Ridge_Regression
gptkbp:effectOnCoefficients Sparsity
https://www.w3.org/2000/01/rdf-schema#label Lasso Regression
gptkbp:hyperparameter Regularization strength (lambda)
gptkbp:implementedIn gptkb:MATLAB
gptkb:scikit-learn
R
gptkbp:introduced gptkb:Robert_Tibshirani
gptkbp:introducedIn 1996
gptkbp:limitation Can be unstable with correlated predictors
Can select at most n variables if n < p
gptkbp:objective Least squares with L1 regularization
gptkbp:penalty L1 penalty
gptkbp:reduces Sum of squared errors plus L1 norm of coefficients
gptkbp:relatedTo Subset selection
Shrinkage methods
gptkbp:shrinksCoefficients Yes
gptkbp:solvedBy Coordinate descent
Least angle regression
gptkbp:usedFor Regularization
Feature selection
gptkbp:usedIn gptkb:Machine_Learning
Statistics
gptkbp:bfsParent gptkb:Linear_Model
gptkbp:bfsLayer 5