Evolutionary systems & genetic algorithms
From GenerativeArt
(Difference between revisions)
(→Types of Genotype Representations: Improved 4 types of representations based on my papers of late 2009 and early 2010) |
|||
| - | == Types of | + | == Types of Genetic Representations == |
| - | There are at least | + | There are at least 4 types of genotype representation that can be differentiated based on the mechanisms they use. Each has a different complexification capacity, i.e. some representations lead to more complexity than others. |
| - | + | ||
| - | + | ||
| - | + | ||
| + | * '''''Fixed Parametric''''' - A genotype that uses a fixed number of parameters that map into phenotypical characteristics in a one-to-one manner. The complexification capacity of this system is highly constrained. | ||
| + | **For example, consider a system for creating drawings of insects. There might be a gene for head size, another for body color, another for leg length, and so on. While such a system may draw a wide variety of insects it will never draw a spider because unless there is a “number of legs” gene all results will have six legs. | ||
| + | * '''''Extensible Parametric''''' - A slightly more complicated genotype that uses a variable number of parameters that map into phenotypical characteristics in a one-to-one manner. The complexification capacity of this system is still fairly constrained. | ||
| + | **An extension of the previous example might allow an arbitrary number of genes for legs. By allowing each gene to draw a single leg this system would be able to draw insects, spiders, and even centipedes and millipedes. But it would not be capable of drawing fish or birds because it lacks fin and wing genes. | ||
| + | * '''''Direct Mechanical''''' - This type of genotype describes one or more machines that in turn construct the phenotype. The complexification capacity of this kind of system is potentially much greater than the previous two. | ||
| + | **In our example this genetic system doesn’t describe a drawing, but rather describes a machine that can draw. Such a representation will, in theory, allow most anything to be drawn. In addition, during reproduction the genes themselves may mutate making the child different than the parent. For example, a machine that creates thin pencil lines may mutate into a machine that makes brushed ink marks. Such a system may seem to be of unlimited potential, i.e. unlimited complexification capacity. But such a system is only capable of a single layer of emergence. The machines immediately and directly draw the picture, and that is that. | ||
| + | * '''''Reproductive Mechanical''''' - Such a system is similar to the previous one, with the significant addition that within a single individual a machine may also create another machine, reproduce itself, or contribute to an emergent machine at a higher level of complexity and scale. This genetic representation offers the greatest complexification potential. | ||
| + | **And this is, in fact, the kind of genetic representation found in nature. There is an upwardly layered increase of complexity as DNA creates proteins, proteins organize to create organelles, organelles organize to create cells, cells organize to create organs, and so on. | ||
== Practical Notes for Computer Artists == | == Practical Notes for Computer Artists == | ||
