Automated prediction of preference level by artificial neural network for simple geometry

This paper provides an overview and analysis of research-in-progress and thought-provoking work on user preference simulation-based design tool. This tool automates prediction of aesthetic preference levels for a simple abstract geometry. An artificial neural network is trained on a random sample of interactively user-evaluated geometries and creates a user preference profile. This profile is then used to automatically generate a new geometry that would theoretically be preferred by all users with different profiles. This saves time and avoids user fatigue compared to interactive genetic evolution. The tool is implemented as an online tool running in a web-browser. In this paper we will present the relationship between the artificial neural network and the genetic algorithm that are used for interactive creation of a preference user profile to stimulate further research in the area of preference level prediction and automation.