In this paper, we develop techniques based on evolvability statistics of the fitness land-scape surrounding sampled solutions. Averaging the measures over a sample of equal fitness solutions allows us to build up fitness evolvability portraits of the fitness land-scape, which we show can be used to compare both the ruggedness and neutrality in a set of tunably rugged and tunably neutral landscapes. We further show that the tech-niques can be used with solution samples collected through both random sampling of the landscapes and online sampling during optimization. Finally, we apply the techniques to two real evolutionary electronics search spaces and highlight differences between the two search spaces, comparing with the time taken to find good solutions through search.
MO supervised Tom Smith (my postgraduate student) and provided model system for exploring the evolvability of novel styles of artificial neural networks inspired by gaseous signalling by NO. Originality: Developed new mathematical and computational techniques for characterising evolvability in a fitness landscape. Rigour: A set of continuous metrics were developed which allows a much more detailed characterisation than with previous global measures. These were applied to tuneable model landscapes and to real landscapes associated with evolutionary electronics problems. Significance: The novel methods developed enable the construction of evolvability portraits based on local characteristics of the fitness landscape surrounding a solution. These can be used to characterise properties such ruggedness and neutrality that are highly pertinent to search efficiency. Impact: Tools developed are becoming increasingly widely used as they allow the construction of fitness evolvability portraits for real problems. 43 Google Scholar citations, 13 Web of Science.