Towards a simulator dedicated to solving optimization problems using Potts neural networks

A general library has been created for the development and test of applications of the Potts neural network paradigm to optimization problems. As more applications of Potts neural networks are being developed the need for a dedicated simulator is growing. Most neural network simulators are designed to develop and simulate neural networks with learning. The basic algorithm is compact and could be implemented with each application. The author found that it was still worth the effort to develop a library which separates the basic algorithm from the application code. Development of new applications can be faster and further optimizations to the simulation code will apply to all applications. Also it will be easier to port applications to specialized hardware. Design decisions are justified with references to the applications developed. Graph coloring is used as an example application. The library has been implemented in C++