Vector quantization by deterministic annealing
Kenneth H. Rose, Eitan Gurewitz, Geoffrey Fox · IEEE Transactions on Information Theory · 1992
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.>
Cited by 262 works. doi:10.1109/18.144705
Authors
References
- A Mathematical Theory of Communication
- A Mathematical Theory of Communication
- Optimization by Simulated Annealing
- Equation of State Calculations by Fast Computing Machines
- Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images
- Information theory and statistics
- Self-Organization and Associative Memory
- Information Theory and Statistical Mechanics
- An Algorithm for Vector Quantizer Design
- Least squares quantization in PCM