Exploring the Impacts of Pseudo-Random Number Generators on Sub-pixel Classification

Pseudo-random number generators (PRNGs) create deterministic approximations of random number sequences that typically contain biases and structural correlation. While this non-random behavior is relatively well known, its impact on geographic applications has received little attention in the geoscie...

Full description

Saved in:
Bibliographic Details
Published inGIScience and remote sensing Vol. 45; no. 4; pp. 471 - 489
Main Authors Makido, Yasuyo K., Messina, Joseph P., Shortridge, Ashton M.
Format Journal Article
LanguageEnglish
Published Taylor & Francis Group 01.10.2008
Online AccessGet full text

Cover

More Information
Summary:Pseudo-random number generators (PRNGs) create deterministic approximations of random number sequences that typically contain biases and structural correlation. While this non-random behavior is relatively well known, its impact on geographic applications has received little attention in the geoscience literature. This study examines the role of specific PRNG algorithms and initializations on a super-resolution (sub-pixel) map classification technique, the simultaneous categorical swapping method. The results indicate that the choice of PRNG as well as the specific seed number significantly impact the sub-pixel allocation model. The paper concludes by considering the broader implications of propagation of non-random PRNG behavior for geospatial applications.
ISSN:1548-1603
1943-7226
DOI:10.2747/1548-1603.45.4.471