Genetic programming, evolutionary algorithms

Description:

A population of programs is evaluated according to a fitness function before operators like cross-over and mutation change selected individuals, which gives us the new population. The method can be seen as heuristic search in the space of all possible programs. Hence, the programs must be restricted in some kind.

Publications: Baeck/etal/97a: Evolutionary computation: Comments on the history and current state
Esquivel/etal/2002a: Enhanced evolutionary algorithms for single and multi-objective optimization in the job shop scheduling problem
Holland/86a: Escaping Brittleness: The Possibilities of General--Purpose Learning Algorithms Applied to Parallel Rule--Based Systems
Macready/Wolpert/98a: Bandit problems and the exploration/exploitation tradeoff.