Formulation of Aesthetic Evaluation & Selection

Abstract

Abstract. The application of evolutionary algorithms in an aesthetic domain involves human subjective judgment and artificial selection. Evolutionary systems that facilitate this selection strategy are referred to as Interactive Evolutionary Systems (IES). One of the crucial problems for IES is that artificial selection is a time consuming process during which human users are faced with limited scale of the population and high dimensionality of solution space. This paper addresses this problem through an integration of General Regression Neural Network (GRNN) and an IES, using facial character creation as an example domain. This approach formulates designers’ aesthetic fitness evaluation in an IES through a learning mechanism provided by the GRNN. Our aim is to build an intelligent and evolutionary system that possesses some empirical knowledge as well as a convergent thinking ability to support human users. In order to study the feasibility of our approach, we have implemented a prototype system for evolutionary facial character creation. The initial results generated by this system are reported in this paper.