Full text
| |
| Author(s): |
Ernesto Cuadros Vargas
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | São Carlos. |
| Institution: | Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB) |
| Defense date: | 2004-03-03 |
| Examining board members: |
Roseli Aparecida Francelin Romero;
André Carlos Ponce de Leon Ferreira de Carvalho;
Paulo Martins Engel;
Agma Juci Machado Traina;
Fernando José von Zuben
|
| Advisor: | Roseli Aparecida Francelin Romero |
| Abstract | |
largo and complex Data Bases. Two groups of techniques are widely used for it, Self-Organizing Maps (SOM) and - Spatial Access Methods (SAM) and Metric Access Mothods (MAM). Howovor. l)oth groups of techniques present iniportant drawbacks. Most of SOM make intensivo use1 of sequential comparison to find winner units. On the other hand, Access Mothods do not take advantage of knowledge generated by previous queries. In order lo overeome t liese problenis. two novel techniques are proposed to improve the SIH process. The lirst technique is based 011 SOM and the second one 011 SAM and MAM. Firstly. SOM was used jointly with SAM and MAM in order to improve SOM based svstenis. produoing two now families of techniques narned SAMSOM and MAMSOM rospoet i volv. Seoondlv. SAM and MAM themselves were improved by the creation of PMAM, a phig-111 modulo whieh is iiseful to take advantage of knowledge acquired by previous queries 111 order lo speed up following queries. The coinbination of PMAM jointly with SAM and MAM produced MAE+ and MAM+ families of Access Methods. As the niaiii result. SAM-SOM and MAM-SOM families outperform considerably traditional SOM based systems which norrnally need a long time for training. Aditionally. MAE+ and MAM+ families are capable of reducing gradually the nuniber of operations needed to answer a query, as new queries are introduced. This is possible because PMAM allows thern to take advantage of knowledge generated by sucessivo queries. (AU) | |