Visualising evolutionary search spaces
James McDermott · ACM SIGEVOlution · 2014
Understanding the structure of search spaces can help us to design better search algorithms, and it is natural to try to understand search spaces by visualising them. For typical evolutionary search spaces, like the space of genetic programming trees, visualising them directly is impossible, because of their large dimensionality. However, we can use the idea of distances on search spaces to project them into two dimensions, expose their structure, and obtain useful and attractive visualisations.