Partial Breeding - a method of IEC for well-structured large scale target domains
Tatsuo Unemi · 2003
This paper proposes a method of Interactive Evolutionary Computation (IEC) for large scale target domains of which structure is well organized. In general, It is effective to divide a large problem into a number of sub-problems to solve it efficiently. For IEC tools, it becomes possible by adding a GUI-based facility for the user to indicate some part of genotype protected against random Initialization and mutation. The user can try alternative candidates only on the non-protected parts as a sub-problem through breeding process: iteration of subjective selection, mutation, and crossover. This method was invented through development of an application for musical composition support system, but it is useful for other domains, such as graphics.