Spelling suggestions: "subject:"trianguliacija"" "subject:"populiacijos""
1 |
Daugiamačių duomenų projekcijos algoritmai, mažinantys atstumų skaičiavimus / Algorithms of the multidimensional data projection, decreasing calculations of distancesKarbauskaitė, Rasa 10 June 2005 (has links)
In this work, a triangulation method, a classic algorithm for Sammon’s projection and the combination of Sammon’s algorithm and the triangulation method for mapping of new points are examined in details. The combination of Sammon’s algorithm and triangulation method is created. The algorithms of triangulation method, Sammon’s algorithm and their combination are realized in Microsoft Visual Basic 5.0. These algorithms are developed and examined on the view on the following stand points:
* Visual evaluation of data projection;
* Evaluation of time of data mapping;
* Evaluation of projection error.
The triangulation method is quite fast, however only (2m-3) distances among examined data points are exactly preserved. Sammon’s algorithm tries to preserve all m(m-1)/2 distances among data points, but it is quite slow: whenever a new point must be mapped, all visualization procedure must be repeated over.
This question is solved using the combination of the triangulation method and Sammon’s projection. On purpose to get the more exact data projection onto a plane, we must to map more initial points using Sammon’s projection and less new points using the triangulation method. The combination of these methods is used, whenever new points of the data must be quickly mapped. This method performs quite fast and quite a little precision is lost.
The discovered new ways for fast mapping of points in a high-dimensional space onto a plane and it makes a background for the further research... [to full text]
|
Page generated in 0.0343 seconds