Boosting Interactive Evolution Using Human Computation Markets

摘要: Interactive evolution, i.e. leveraging human input for selection in an evolutionary algorithm, is effective when appropriate fitness function hard to quantify yet solution quality easily recognizable by humans. However, single-user applications of interactive evolution are limited user fatigue: Humans become bored with monotonous evaluations. This paper explores the potential bypassing such fatigue directly purchasing from computation markets. Experiments evolving aesthetic images show that purchased can be leveraged more economically first seeded optimizing a purely-computational measure. Further experiments same domain validate system feature, demonstrating how help guide design. Finally, image composition approach’s make scalable even tasks not inherently enjoyable. The conclusion markets it possible apply powerful form pressure mechanically algorithms.