Sonia1234
Junior Member level 2
- Joined
- Sep 25, 2011
- Messages
- 21
- Helped
- 0
- Reputation
- 0
- Reaction score
- 0
- Trophy points
- 1,281
- Activity points
- 1,446
Hello,
I've been going through Linde-Buzo Gray algorithm (1980) and Lloyd's k-means algorithm and both seems to be doing exactly the same thing (or maybe I'm not researching well).
They seem to be going through the following:
1. compute initial codebook randomly from training set.
2. select training vectors that are closest to each codeword within codebook (using Euclidean distance measure)
3. once vectors belonging to each codeword are selected, compute their average and replace it with the old codeword.
4. do this for all codewords within the codebook
5. once the codebook is updated, find the distortion between old and the new codebook
6. if distortion between the two codebooks is less than pre-defined threshold, terminate the algorithm. Otherwise, repeat.
I would appreciate if someone with more knowledge in this area could point out the main difference between the two algorithms.
Thank you
I've been going through Linde-Buzo Gray algorithm (1980) and Lloyd's k-means algorithm and both seems to be doing exactly the same thing (or maybe I'm not researching well).
They seem to be going through the following:
1. compute initial codebook randomly from training set.
2. select training vectors that are closest to each codeword within codebook (using Euclidean distance measure)
3. once vectors belonging to each codeword are selected, compute their average and replace it with the old codeword.
4. do this for all codewords within the codebook
5. once the codebook is updated, find the distortion between old and the new codebook
6. if distortion between the two codebooks is less than pre-defined threshold, terminate the algorithm. Otherwise, repeat.
I would appreciate if someone with more knowledge in this area could point out the main difference between the two algorithms.
Thank you