Chapter 20. K-means Clustering

Lioyd Max Quantization

Quantization is a kind of clustering

• Data coding and compression - Quantization : codewords
• Pattern recognition - Clustering : Representative patterns

동작 과정

1. Select K initial centroid vector(좌표)
   • Random or Simple rules
2. Find the closest vectors for every centroid using Euclidean distances
   • Construct a tentative cluster for every controid
3. Update the centroid(=mean) of every cluster
4. Repeat 2 and 3 processes until stop conditions
   • little change of centroids
   • Maximum iteration number