Implementace top-trees

Abstract

Nazev prace: Implerneiitace top-trees Autor: Jifi Sofka Katedra (ustav): Katedra teoreticke informatiky a matematicke logiky Vedouci diplornove prace: Mgr. Vladan Majerech, Dr. e-mail vedouciho: Vladan.Majerech@mff.cuni.cz Abstrakt: Top Trees reprezentuji puvodni strom pomoci jeho dekompozice do binarni hierarchie kontrakcnich clusteru. Kazdy z techto clusteru obsahuje agregovana data o celem podstromu, jehoz je kofenem. Mezi pfednosti Top Trees patfi oddeleni spravy dat clusteru do funkci Create, Destroy, Join, Split a Select, ktere jsou automaticky volany pfi kazde zmene hierarchie clusteru. Diky tomu je irnplementace Top Trees zcela nezavisla na datech, ktera uchovava. Velikost dat agregovanych v clusteru, ktere pfislusi k jedne udrzovane vlastnosti o podstromu, muze byt zavisla na velikosti celeho reprezento- vaneho stromu. Stejne tak jejich vypocet muze byt casove narocny. Pfi dota- zovani Top Trees se casto pouzivaji operace, ktere zpusobi zmenu struktury clusteru. V puvodnim navrhu Top Trees je pfi kazde takove zmene nutna ak- tualizace dat vsech sledovanych vlastnosti ve zmenenych clusterech, nehlede na to, zda jsou vsechny tyto vlastnosti relevantni k polozenemu dotazu. Ak- tualizace dat nerelevantnich vlastnosti je zpravidla zbytecna. Tato prace detailne popisuje implementaci Self-Adjusting Top...

Title: Top Trees implementation Author: Jifi Sofka Department: Department of Theoretical Computer Science and Mathemat- ical Logic Supervisor: Mgr. Vladan Majerech, Dr. Supervisor's e-mail address: Vladaii.Majerech@mff.cmii.cz Abstract: Dynamic data structure Top Trees is used to represent a tree by a binary structure of tree contractions. Vertices of such a binary tree are called clusters. Each cluster represents one contraction and contains aggregated data of its children. One of the main advantages of Top Trees is the separation of cluster's data management into Create, Destroy, Join, Split and Select functions which are automatically called as the internal structure of Top Trees changes. The implementation of Top Trees is therefore independent of all the properties that are tracked about the original tree (i.e. 2-connecticity, tree median). The size of cluster's aggrefated data about one tree property may de- pend on the size of the entire original tree and its calculation can be time consuming. The operations used to query the Top Tree about different tree properties often result in Top Tree structural changes. In the original Top Tree design, each of the affected clusters needs to update the data belong- ing to all tracked properties when such structural change occurs. Not all of the properties are...