Refine
H-BRS Bibliography
- yes (1) (remove)
Departments, institutes and facilities
Document Type
Year of publication
- 2015 (1) (remove)
Language
- English (1)
Has Fulltext
- no (1)
Keywords
- HyperNEAT (1) (remove)
A novel approach to produce 2D designs by adapting the HyperNEAT algorithm to evolve non-uniform rational basis splines (NURBS) is presented. This representation is proposed as an alternative to previous pixel-based approaches primarily motivated by aesthetic interests, and not designed for optimization tasks. This spline representation outperforms previous pixel-based approaches on target matching tasks, performing well even in matching irregular target shapes. In addition to improved evolvability in the face of a well defined fitness metric, a NURBS representation has the added virtues of being continuous rather than discrete, as well as being intuitive and easily modified by graphic and industrial designers.