Optimization with Potts Neural Networks
The Potts Neural Network approach to non-binary discrete optimizationproblems is described. It applies to problems that can be described asa set of elementary 'multiple choice' options. Instead of the conventionalbinary (Ising) neurons, mean field Potts neurons, having several availablestates, are used to describe the elementary degrees of freedom of suchproblems. The dynamics consists of iteratin
