Quality-diversity in dissimilarity spaces

We apply this framework space (which need not span the entire space, since we can use to formulate quality-diversity algorithms in generic dissimilarity the output of one run of the algorithm to initialize another); spaces. In particular, we instantiate a very general version of Go-an efficient mechanism for locally perturbing existing points; Explore with promising performance for challenging and computationally and a mechanism for estimating the objective that permits expensive objectives, such as arise in simulations. Finally, efficient evaluation: e.g., interpolation using polyharmonic we prove a result on diversity at scale zero that is interesting in its radial basis functions [19] or a neural network.