Vector quantization by deterministic annealing |
| |
Authors: | Rose K. Gurewitz E. Fox G.C. |
| |
Affiliation: | Dept. of Electr. & Comput. Eng., California Univ., Santa Barbara, CA; |
| |
Abstract: | A deterministic annealing approach is suggested to search for the optimal vector quantizer given a set of training data. The problem is reformulated within a probabilistic framework. No prior knowledge is assumed on the source density, and the principle of maximum entropy is used to obtain the association probabilities at a given average distortion. The corresponding Lagrange multiplier is inversely related to the `temperature' and is used to control the annealing process. In this process, as the temperature is lowered, the system undergoes a sequence of phase transitions when existing clusters split naturally, without use of heuristics. The resulting codebook is independent of the codebook used to initialize the iterations |
| |
Keywords: | |
|
|