Hyperright parallelepiped

Hyperright parallelepipeds (not to confound with hypercubes) position certain characteristic vectors in a n-dimensional area and divide these values determined, before trained classes too. Usually this is done via a n-dimensional Voronoi diagram.

A characteristic vector or also a sample, is a simple vector, which determines characteristics of data contains, for example from a screw the length, thickness, type of thread, etc.

Examples:

1. The characteristic vector contains 2 values: Color red (0 255) and color blue (0 255). By the minimum and maximum values one can already see, in which range combinations to lie to be able. Then one provides a Voronoi diagram and gives the individual cells of values as "“property"” or "“bad"”. The x axis is color red, the y axis color blue. If one has now a combination of the values, one can clearly determine whether this "“property"” or "“bad"” is. An extension could be to determine whether the combination a red, lila one or a blue represents.
2. One introduces oneself a sort box for screws. Sorted according to thickness and length the screws are sorted thereby into the individual fan. X axis is thereby the thickness and y axis is then the length. The characteristic vector contains thereby the characteristics thickness and length. With the respective data can thus be determined into which small box a screw comes. In this example a hyperright parallelepiped 2 with a Voronoi diagram would be dimensional.

The whole can be expanded since any number of characteristics.

One can reach same with Fuzzysystemen and neural networks. A tabular listing of the pro and/or cons would be desirable. From hyperright parallelepipeds also Fuzzyregeln can be derived, whereby these are however not so well suitable to represent a hyperright parallelepiped. Hyperright parallelepipeds can be folded up through to also generalize. This is from use, if one liked to classify many samples. In addition one folds up 2 in each case or more hyperright parallelepipeds and tries first to classify with this the sample.

See also:

• Fuzzy
• Artificial intelligence
• Neural networks
• Klassifikator
• NGE theory (Nested Generalized copy)

Articles in category "Hyperright parallelepiped"

We found here 2 articles.