Unsupervised Learning and Clustering Padhraic Smyth Information and Computer Science ICS 175, Spring 2002

Example: Data in 2 Clusters Feature 2 Feature 1

“Compact” Clustering: Low TSE Feature 2 Cluster Center 2 Cluster Center 1 Feature 1

“Non-Compact” Clustering: High TSE Feature 2 Cluster Center 2 Cluster Center 1 Feature 1

Original Data (2 dimensions)

Initial Cluster Centers for K-means (K=2)

Update Memberships (Iteration 1)

Update Cluster Centers at Iteration 2

Update Memberships (Iteration 2)

Update Cluster Centers at Iteration 3

Update Memberships (Iteration 3)

Update Cluster Centers at Iteration 4

Updated Memberships (Iteration 4)

Clustering Images We can also cluster sets of images into groups now each vector = a full image (dimensions 1 x (mxn)) N images of size m x n convert to a matrix with N rows and (m x n) columns just use image_to_matrix.m call kmeans with D = this matrix kmeans is now clustering in an (m x n) dimensional space kmeans will group the images into K groups

Example: First 5 Individuals, K = 2 Cluster 1 Cluster 2

Example: 2nd 5 individuals, K = 2 Cluster 1 Cluster 2

All Individuals, Happy Faces, K=5