The UNN optimization problem is particularly difficult to solve due to local optima. The two iterative strategies UNN and UNNgof the last chapters allow the fast construction of a solution. In this chapter, we want to answer the question, if exhaustive evolutionary search allows to come closer to the optimal embedding. We compare a discrete evolutionary approach based on stochastic swaps to a continuous evolutionary variant that is based on evolution strategies, i.e., the covariance matrix adaptation variant CMA-ES. The continuous variant is the first step to embeddings into continuous latent spaces.
CITATION STYLE
Kramer, O. (2013). Metaheuristics (pp. 75–91). https://doi.org/10.1007/978-3-642-38652-7_6
Mendeley helps you to discover research relevant for your work.