A novel human-computer collaboration: combining novelty search with interactive evolution

摘要: Recent work on novelty and behavioral diversity in evolutionary computation has highlighted the potential disadvantage of driving search purely through objective means. This paper suggests that leveraging human insight during can complement such novelty-driven approaches. In particular, a new approach called novelty-assisted interactive (NA-IEC) combines intuition with to facilitate serendipitous discovery agent behaviors deceptive maze. this approach, user directs evolution by selecting what is interesting from on-screen population behaviors. However, unlike typical IEC, now request next generation be filled novel descendants. The experimental results demonstrate combining not only finds solutions significantly faster at lower genomic complexities than fully-automated processes guided fitness or novelty, but it also traditional IEC approach. Such add evidence users automated creates synergistic effect for solutions.