Illustration of the LGB Vector Quantization Algorithm
This presentation has been prepared with
webslider.
The slides may be
perused by selecting any of the entries in this Table of Contents.
- Training vectors as points in a 2D space
- First step, the centroid of the entire set
- Split the centroid into two (randomly)
- Nearest-neighbor search, result in 2 clusters
- Update the centroid from each cluster
- Repeat nearest-neighbor search
- And repeat centroid update
- Until the centroids are stable (end of the inner loop)
- Repeat outer loop, double the number of centroids by splitting
- Iterate inner loop until the centroids are stable
- Continue...
- 8 centroids
- And...
- Finally, stop when the desired number of centroids is obtained
Copyright © Minh N. Do 2000