Improving Evolvability through Generative Representations

Abstract

One of the main limitations of computer automated design systems is the representation used for encoding designs. Using computer programs as an analogy, representations can be thought of as having the properties of combination, control-flow and abstraction. Generative representations are those which have the ability to reuse elements in an encoding through either iteration, a form of control-flow, or abstraction. Here we argue that generative representations improve the evolvability of designs by capturing design dependencies in a way that makes them easier to change, and we support this with examples from two design substrates.