k-means

GPTKB entity

Statements (50)
Predicate Object
gptkbp:instanceOf gptkb:algorithm
gptkbp:alternativeName gptkb:Lloyd's_algorithm
k-means clustering
gptkbp:application image segmentation
document clustering
market segmentation
vector quantization
gptkbp:category partitioning method
gptkbp:complexity O(nkt)
gptkbp:convergesTo local minimum
gptkbp:field gptkb:machine_learning
statistics
data mining
https://www.w3.org/2000/01/rdf-schema#label k-means
gptkbp:input set of data points
gptkbp:introduced gptkb:Stuart_Lloyd
gptkbp:introducedIn 1957
gptkbp:limitation sensitive to outliers
assumes spherical clusters
sensitive to initial centroids
requires k to be specified
gptkbp:measures Euclidean
Manhattan (less common)
gptkbp:notRecommendedFor categorical data
non-globular clusters
gptkbp:objective minimize within-cluster sum of squares
gptkbp:optimizedFor gptkb:Lloyd's_algorithm
gptkbp:output clusters
gptkbp:popularizedBy gptkb:James_MacQueen
gptkbp:popularizedYear 1967
gptkbp:purpose partitioning data into clusters
gptkbp:relatedTo gptkb:DBSCAN
gptkb:Gaussian_mixture_model
hierarchical clustering
gptkbp:requires number of clusters (k)
gptkbp:software gptkb:MATLAB
gptkb:Spark_MLlib
gptkb:scikit-learn
R
gptkbp:step assign points to nearest centroid
initialize centroids
repeat until convergence
update centroids
gptkbp:supportsAlgorithm unsupervised learning
gptkbp:uses Euclidean distance
gptkbp:variant fuzzy c-means
k-medoids
mini-batch k-means
gptkbp:bfsParent gptkb:Bag_of_Visual_Words
gptkbp:bfsLayer 7