Article:

Dataflow parametric modeling environments have become popular asexploratory tools due to them allowing the variational exploration of a design bycontrolling the parameters of its parametric model schema. However, the natureof these systems requires designers to prematurely commit to a structure andhierarchy of geometric relationships, which makes them inflexible when it comesto design exploration that requires topological changes to the parametricmodeling graph. This paper is a first step towards augmenting parametricmodeling systems via the use of machine learning for assisting the user towardstopological exploration. In particular, this paper describes an approach whereLong Short-Term Memory recurrent neural networks, trained on a data set ofparametric modeling graphs, are used as generative systems for suggestingalternative dataflow graph paths to the parametric model under development.