Linear Discriminant Analysis
GPTKB entity
Statements (52)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:algorithm
statistical analysis dimensionality reduction technique |
| gptkbp:alsoKnownAs |
gptkb:LDA
|
| gptkbp:application |
finance
marketing diagnosis face recognition |
| gptkbp:assumes |
independent samples
normal distribution of features equal covariance matrices for all classes |
| gptkbp:category |
supervised learning
|
| gptkbp:differenceFromPCA |
LDA uses class labels, PCA does not
|
| gptkbp:differenceFromQDA |
LDA assumes equal covariance, QDA does not
|
| gptkbp:extendsTo |
Kernel Discriminant Analysis
Multiclass LDA Regularized Discriminant Analysis |
| gptkbp:field |
gptkb:machine_learning
statistics pattern recognition |
| gptkbp:form |
maximizes ratio of between-class variance to within-class variance
|
| https://www.w3.org/2000/01/rdf-schema#label |
Linear Discriminant Analysis
|
| gptkbp:implementedIn |
gptkb:SAS
gptkb:MATLAB gptkb:SPSS gptkb:scikit-learn R |
| gptkbp:input |
feature vectors
labeled data |
| gptkbp:introduced |
gptkb:Ronald_A._Fisher
|
| gptkbp:introducedIn |
1936
|
| gptkbp:limitation |
sensitive to outliers
assumes linear boundaries not suitable for non-Gaussian data |
| gptkbp:numberOfConstituents |
at most (number of classes - 1)
|
| gptkbp:output |
discriminant function
projected data |
| gptkbp:purpose |
find linear combinations of features that best separate classes
|
| gptkbp:relatedTo |
Principal Component Analysis
Fisher's Linear Discriminant Quadratic Discriminant Analysis |
| gptkbp:requires |
class labels
feature matrix |
| gptkbp:type |
linear classifier
generative model |
| gptkbp:usedFor |
gptkb:dictionary
dimensionality reduction feature extraction |
| gptkbp:bfsParent |
gptkb:LFSe
gptkb:Linear_Classifier gptkb:Generalized_Discriminant_Method |
| gptkbp:bfsLayer |
8
|