Abstract/Details

Adaption to nonstationary binary sources for data compression

Zandi, Ahmad. 
 University of California, Santa Cruz ProQuest Dissertations Publishing,  1992. 9224688.

Abstract (summary)

The use of statistical methods in general and sequential analysis and Bayesian viewpoint in particular have been shown to be suitable and appropriate in statistical data compression. Our particular goal is the detection and adaptation to non-stationary swings in binary data. Through theoretical and experimental processes the adaptation rate, as a measure for non-stationary (which is defined in this dissertation) prove to be quite efficient. It measures the adaptation of some known and some new algorithms and provides us with new insight. This insight together with direct application of the methods of sequential analysis and adaptation to the adaptation rate lead to new and most efficient algorithms for compression of binary data. Note that the binary source condition is by no means restrictive. As a result of an axiom of Shannon required by the definition of entropy, an information source defined over any alphabet has an equivalent form defined over the binary alphabet.

Our main contribution to binary data compression is contained in three algorithms; fixed-rate, adaptive-rate and LPS 2-1-2. In the fixed algorithm a fixed and predefined rate was used. In the adaptive algorithm the rate is dynamically adjusted and the improvement shows that the nonstationary can change within (as well as between) conditional contexts. The LPS 2-1-2 algorithm is a simple adapter.

Another application which uses our most general algorithm is tracking the adaptation and fine tuning of algorithms. The adaptation rate, in this process, locates the area that the algorithm adapts well and finds the property of the data that causes it to perform below optimum.

A major contribution of this dissertation has been to establish the fact that the adaptation to nonstationary is at the cost of coding efficiency for stationary data. But this need not to be done heuristically. There are concrete measures for evaluating the loss and the gain and hence resulting in appropriate decisions.

Indexing (details)


Subject
Computer science
Classification
0984: Computer science
Identifier / keyword
Applied sciences; binary data
Title
Adaption to nonstationary binary sources for data compression
Author
Zandi, Ahmad
Number of pages
65
Degree date
1992
School code
0036
Source
DAI-B 53/04, Dissertation Abstracts International
Place of publication
Ann Arbor
Country of publication
United States
ISBN
979-8-208-17299-5
Advisor
Langdon, Glen G., Jr.
University/institution
University of California, Santa Cruz
University location
United States -- California
Degree
Ph.D.
Source type
Dissertation or Thesis
Language
English
Document type
Dissertation/Thesis
Dissertation/thesis number
9224688
ProQuest document ID
303987140
Copyright
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.
Document URL
https://www.proquest.com/docview/303987140