|
| |
|
| |
|
| |
|
|
Jak vypadá prostor kudy kráčí e-Golem?
Jiřina, Marcel ; Jiřina jr., M.
It is shown that in the multivariate data space the distance from particular point changes its nature with dimension. In the data set of reasonable size in high dimension all or nearly all points are influenced by boundary effect. Some features of such observation are drawn and suggestion how to compensate the data spacious deformation is shown.
|
|
| |
|
| |
|
| |
|
| |
|
|
Okrajový jev v mnohorozměrných datech
Jiřina, Marcel ; Jiřina jr., M.
In this paper we show some strange features of multidimensional data and their influence on classification. We introduce the probability distribution mapping function, and the distribution density mapping function which maps probability density distribution of points in n-dimensional space to a similar distribution in one-dimensional space of distances. The power approximation of the probability distribution mapping function is introduced and an application for a probability density estimation including the boundary effect in high dimensions is shown. Some results obtained with the new method for classification are shown.
|
|
| |
guest ::
