Using data sets to simulate evolution within complex environmentsBruce Edmonds
The talk suggests a way to investigate whether environmental complexity affects evolutionary processes. The suggestion is to use complex data sets that derive from the observation/measurement of natural systems. The data set is spread accross a space defined by two of the variables in the data set. The genome of individuals encode a model that tries to predict the outcome variable from the independent variables in the set. Resources in the model are gained by individuals local to a data point predicting the outcome variable better than its competitors. Once the contribution of resources to all individuals is decided, then an evolutionary process takes place, preferentially propogating and mating fitter individuals (so the less fit die). This is illustrated with an evolutionary model using the Cleveland Heart Disease Data Set. Evolution on this is compared to evolution on a data set with similar distributions of values for each variable but chosen independently of each other, so with the noise and shape of the original but without the complexity. Illustrative results showing that the change in environment significantly affects how evolution proceeds is exhibited and discussed. It is suggested that the same approach could be used to test the assumptions (e.g. the adequacy of a simple environment) in many models of evolutions, as well as investingating whe environmental complexity might affect evolution and what kind of complexity.