Gradual Drifting of the User´s Preference in Interactive Evolutionary Computation and its Application

The difference between interactive evolution-ary computation (IEC) and traditional evolutionary computation (TEC) is that in IEC individuals´ fitness is subjectively assigned by the user, while in TEC the fitness is objectively given by function or others. The user in IEC assigns fitness according to his/her preference. Therefore, if his/her preference drifts, the implicit fitness function for preference would also drift. This will cause the invalidation of the history evolutionary information such as individuals fitness and so on. In this situation, it is the key issue that how to make use of the history evolutionary information to accelerate evolving without misleading the evolutionary direction. The gradual drift of the user´s preference is mainly studied in this paper. The definition of gradual drift of the user´s preference is given and the methods to make use of the history evolutionary information are put forward. At last, the experiments show the efficiency of these methods.