> ML_ALGORITHM // K-MEANS-CLUSTERING_v1.0
K-Means & K-Means++ Clustering
The benchmark iterative partitioning algorithm that minimizes intra-cluster Euclidean variance around k centroid prototypes.
Partitioning Clusteringclassical-unsupervisedhigh-intrinsicmedium (1k-100k)
Back to All AlgorithmsComputational Complexity
Training Complexity:O(iterations * k * n * p)
Inference Complexity:O(k * p)
Hardware Profile
CPU Friendly:Yes
Requires GPU:No
Memory Footprint:low
Interpretability & Data
Interpretability Tier:high-intrinsic
Training Data Needs:medium (1k-100k)
Interpretability Assessment
Centroids directly represent the average feature vector of all members in each cluster.
Suitable Tasks & Supported Modalities
Suitable Tasks:
clusteringvector quantization
Supported Modalities:
tabular
Implementing Libraries
Foundational Literature
Least squares quantization in PCMStuart P. Lloyd (1982) · IEEE Transactions on Information Theory
k-means++: The Advantages of Careful SeedingDavid Arthur, Sergei Vassilvitskii (2007) · ACM-SIAM Symposium on Discrete Algorithms (SODA)
Common Pitfalls & Warnings
- Selecting k arbitrarily without elbow curve or silhouette validation
- Failing on non-convex or unevenly sized clusters