Deterministic annealing, constrained clustering, and optimization

In previous work the authors (Phys. Rev. Let., vol.65, p.945-8, 1990) proposed the concept of deterministic annealing for the problem of clustering and vector quantization. This approach is summarized. The authors extend the clustering method to the constraint clustering method. Adding constraints to the deterministic annealing mechanism expands the variety of optimization problems which can be solved by this method. A brief presentation of the clustering approach is given. Two examples to which the constraint clustering approach can be applied are included.