Art and analogy through evolutionary computation

Abstract

In this thesis the development of a computer system capable of generating art through the process of analogy is presented. A visual display is generated analogous to a given piece of music. This analogy is based upon the aesthetics of both music and visuals whereby an aesthetic piece of music should prompt an aesthetic visual display with similar characteristics.

The aesthetics of both music and visuals which form the core of the proposed system remain difficult to capture. A novel approach is presented, designed to capture and evaluate aesthetic experience for use in computational systems using a graph based ranking approach. A number of experiments were conducted to analyse the performance of this approach, and a number of studies were conducted using this approach to gather aesthetic data.

Representing an analogy within a computer system requires codifying the connections between two domains. Mapping expressions are introduced to represent these connections. Together with the aesthetic data gathered, these mapping expressions are evolved using a computational evolution approach. The output of this process is an abstract representation of an analogy. This representation is then used to create a visual display.

Further studies have been carried out to evaluate the effect of these generated visuals on the aesthetic experience of a viewer.