Soft Computing Techniques
This chapter provides a brief description of optimization and soft computing techniques. It presents a brief introduction to: genetic algorithms, linear programming, simulated annealing, tabu search, terminal repeller unconstrained subenergy tunneling, artificial neural networks and fuzzy sets. Artificial neural networks, like simulated annealing and genetic algorithms, originated as a simulation of a natural phenomena. Neural networks are also called connectionist learning schemes, since they store information in weights given to connections between the artificial neurons. Neural networks can also be implemented to serve as self-associative memories. The network used for this type of application generally consists of a single layer of neurons, all of which are fully connected. Training sets that are too large can result in over-training the network so that the network does not correctly generalize the problem and only recognizes the instances used in training the network.