Efficient graph based representation of multi-valued functions with an application to genetic algorithms

We present a new general data structure for representation of multi-valued input, multi-valued output functions, called function graphs (FG). Ordered FGs are shown to be a canonical form. We investigate read-once FGs and show that type-restricted FGs are also canonical. We use FGs for multi-valued set representation and manipulation. They allow efficient manipulation algorithms for set operations, e.g. union and intersection. An application to genetic algorithms (GAs) is presented in more detail. A population is represented by a multi-rooted FG. It is shown by experimental results that this new representation is very efficient and superior to other data structures for GAs.<>