Techniques for More Productive Genetic Design: Exploration With GAs Using Non-Destructive Dynamic Populations (2013)
article⁄Techniques for More Productive Genetic Design: Exploration With GAs Using Non-Destructive Dynamic Populations (2013)
abstract⁄The products of generative design are ever more commonly explored and refined through evolutionary search techniques. Genetic algorithms GAs belong to this class of stochastic procedures, and are particularly wellsuited to the way designers investigate a problem. GAs search by mixing and matching different parts of a solution, represented as parametric variables, to find new solutions that outperform their predecessors. Generally the method proceeds through generations of populations in which the better solutions outsurvive their less desirable siblings. Inherent to this approach, however, is the fact that all but the select solutions perish. This paper discusses a nondestructive GA that uses dynamic populations drawn from a bottomless pool of solutions to find the most productive breeding pairs. In a typical GA the survival or destruction of a solution depends on a welldefined fitness function. By not enforcing the destruction of less fit individuals, the possibility is held open to modify the fitness function at any time, and allow different parts of the solution space to be explored. This ability is ideal for more complex multiobjective problems that are not easily described by a single fitness function. Generally, design presents just such a problem.
|
|
| Year |
2013 |
| Authors |
Von Buelow, Peter. |
| Issue |
ACADIA 13: Adaptive Architecture |
| Pages |
227-234 |
| Library link |
N/A |
| Entry filename |
techniques-more-productive-genetic-design |
