Optimization and Expert Systems with Neural Networks

Neural Networks (NN) provide the framework for the optimization of highly complex problems, known as NP-complete. At the same time NN allow in an elegant way for the implementation of forward and backward chaining Expert Systems (ESs) where the knowledge is represented by production rules but non-explicit domain knowledge can also be learnt. The use of fuzzy logic allows for the processing of partial and uncertain information. As a representative example for optimization we discuss the Traveling Salesman problem (TSP) covering also recent progress, and subsequently we focus on the connectionist ESs, some of them using fuzzy logic. We finally point to the possibility of a unified framework.