Static field approach for pattern classification

Dymitr Ruta, Bogdan Gabrys

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Recent findings in pattern recognition show that dramatic improvementfof the recognition rate can be obtained by application of fusion systemsfutilizing many different and diverse classifiers for the same task. Apart from afgood individual performance of individual classifiers the most important factorfis the useful diversity they exhibit. In this work we present an example of afnovel non-parametric classifier design, which shows a substantial level of diversityfwith respect to other commonly used classifiers. In our approach inspirationffor the new classification method has been found in the physical world. Namelyfwe considered training data as particles in the input space and exploited the conceptfof a static field acting upon the samples. Specifically, every single datafpoint used for training was a source of a central field, curving the geometry offthe input space. The classification process is presented as a translocation in thefinput space along the local gradient of the field potential generated by the trainingfdata. The label of a training sample to which it converged during the translocationfdetermines the eventual class label of the new data point. Based on selectedfsimple fields found in nature, we show extensive examples and visual interpretationsfof the presented classification method. The practical applicabilityfof the new model is examined and tested using well-known real and artificialfdatasets.

Original languageBritish English
Title of host publicationSoft-Ware 2002
Subtitle of host publicationComputing in an Imperfect World - 1st International Conference, Soft-Ware 2002, Proceedings
EditorsDavid Bustard, Weiru Liu, Roy Sterritt
PublisherSpringer Verlag
Number of pages15
ISBN (Print)354043481X, 9783540434818
StatePublished - 2002
Event1st International Conference on Computing in an Imperfect World, Soft-Ware 2002 - Belfast, United Kingdom
Duration: 8 Apr 200210 Apr 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference1st International Conference on Computing in an Imperfect World, Soft-Ware 2002
Country/TerritoryUnited Kingdom


Dive into the research topics of 'Static field approach for pattern classification'. Together they form a unique fingerprint.

Cite this