@inproceedings {pub4005,
	title = {On the Efficiency of a Point Cloud Autoencoder as a Geometric Representation for Shape Optimization},
	author = {Thiago de Jesus de Araujo Rios AND Thomas B{\"a}ck AND Bas van Stein AND Bernhard Sendhoff AND Stefan Menzel},
	year = {2019},
	month = {December},
	abstract = {A crucial step for optimizing a system is the formulation of the objective function, although it is often overlooked, and part of it concerns the selection of the design parameters. One of the major concerns regarding the parameterization is the trade-off between exploring feasible solutions in the design space and maintaining the computational effort at an admissible level, aspect which is closely related to the dimensionality of the problem. In order to achieve such balance in optimization problems with CAE (Computer Aided Engineering) models, the traditional geometric description using constructive approaches can be substituted by deformation methods, e.g. Free Form Deformation, where the position of a few control points might be capable of handling large scale shape modifications. However, in light of the recent developments of Geometric Deep Learning techniques, auto-encoders have risen as a promising efficient alternative for condensing high-dimensional models into simpler representations and to automatically abstract geometric features, taking over part of the user{\textquoteright}s task. Hence, this paper provides an insight on the applicability of the latent space of a Point Cloud Auto-encoder (PCAe) in evolutionary optimization problems with geometric models, by contrasting the responses obtained with a deformational approach to the behavior of the optimizer when the latent space is adopted as representation. Focusing on engineering applications, a target shape matching optimization is used as a surrogate problem due to its similarity to many design tasks, and not only the final result of the experiments are accounted, but also the feasibility of intermediate shapes for CAE analyses is discussed.},
	publisher = {IEEE},
	booktitle = {2019 IEEE Symposium Series on Computational Intelligence (SSCI)},
	city = {Xiamen},
	pages = {791-798}
}