Ordinal versus metric gaussian process regression in surrogate modelling for CMA evolution strategy

This work presents an ordinal-based Gaussian process surrogate model for the state-of-the-art continuous black-box optimizer CMA-ES in scenarios where the objective evaluations are very expensive. Such model is motivated by the CMA-ES' invariance with respect to order preserving transformations. Alongside with the model's description, comparison with the standard (metric) Gaussian process surrogate for the CMA-ES is given.