A service provided by the WU Library and the WU IT-Services

Statistical data compression by optimal segmentation. Theory, algorithms and experimental results.

Steiner, Gottfried (1999) Statistical data compression by optimal segmentation. Theory, algorithms and experimental results. Doctoral thesis, WU Vienna University of Economics and Business.

[img]
Preview
PDF
Download (2259Kb) | Preview

Abstract

The work deals with statistical data compression or data reduction by a general class of classification methods. The data compression results in a representation of the data set by a partition or by some typical points (called prototypes). The optimization problems are related to minimum variance partitions and principal point problems. A fixpoint method and an adaptive approach is applied for the solution of these problems. The work contains a presentation of the theoretical background of the optimization problems and lists some pseudo-codes for the numerical solution of the data compression. The main part of this work concentrates on some practical questions for carrying out a data compression. The determination of a suitable number of representing points, the choice of an objective function, the establishment of an adjacency structure and the improvement of the fixpoint algorithm belong to the practically relevant topics. The performance of the proposed methods and algorithms is compared and evaluated experimentally. A lot of examples deepen the understanding of the applied methods. (author's abstract)

Item Type: Thesis (Doctoral)
Keywords: statistics / data compression / classification / statistical data compression / optimal segmentation / experimental results / optimization / fixpoint algorithm
Divisions: Departments > Finance, Accounting and Statistics > Statistics and Mathematics
Depositing User: Repository Administrator
Date Deposited: 03 Apr 2002 11:12
Last Modified: 15 Sep 2010 00:50
URI: http://epub.wu.ac.at/id/eprint/1939

Actions

View Item