Discrete fitness values for improving the human interface in an interactive GA
Hideyuki Takagi, Kimiko Ohya · 2002
This paper proposes two methods of inputing fitness values that reduce the psychological burden of interactive genetic algorithm operators. The first proposal is an input method that allows discrete fitness values, and that requires less attention from operators in scoring slightly different individuals. The second proposal is an input method that allows both discrete and continuous fitness values. These proposed methods are evaluated using two subjective tests for the task of drawing a face. The results of the subjective tests show that: (1) the first proposed method is significantly better than the method which forces oprators to input only continuous fitness values (p < 0.01), and (2) the second proposed method is significantly better than the first proposed method (p < 0.01).