Computer Science 4442A/B Lecture Notes - Lecture 16: Feature Vector, Maxima And Minima, Unsupervised Learning
8 Mar 2016
School
Department
Course
Professor
Document Summary
4 feature vectors for clustering based on color and image coordinates. K-means clustering: objective function: probably the most popular clustering algorithm, probably the most popular clustering algorithm. K-means clustering: objective function, the most popular clustering algorithm, assumes know the number of clusters should be k, assumes know the number of clusters should be k, optimizes the following function: k x i. 1: optimizes (approximately) the following objective function, optimizes (approximately) the following objective function. Initialization step: pick k cluster centres randomly, algorithm, pick k cluster centers randomly. K-means clustering: algorithm: assign each sample to closest centre, pick k cluster centers randomly, assign each sample to closest center. Iteration step: compute means in each cluster, re-assign each sample to closest mean, iterate until clusters stop changing. K-means: approximate optimization: pick k cluster centers randomly, assign each sample to closest center, k-means is fast and works quite well in practice, but can get stuck in a local minimum of objective jsee.
