Institutional Repository of Vilnius University
Tiesinis neuroninių tinklų kombinavimas: duomenų balinimo operacijos poveikis tiesiniam neuroninio tinklo kombinavimui su įvairaus skirtingumo ekspertais /
Abstract
dc:descriptionThe purpose of this study was to research the effect of data whitening for different linear combination and diversity measure methods selection in context of building better classifier ensemble. In work main characteristics, architecture and combination methods of neural network were identified. Different linear combination methods and diversity measures were analyzed in context of data whitening. In experiment classifier quality effect of data whitening transformation was researched with different classifiers. The new method for designing bad quality different and good quality similar classifiers was constructed. Most important experiment conclusions: -Data whitening transformation applied for initial data vectors makes classifiers more similar. -Applying data whitening transformation for single classifier set answers makes bad quality different classifiers better than good quality similar classifiers in lower generalization error range up to 40%.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2014
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gruodis, Šarūnas,
- Contributors dc:contributor
-
- Janeliūnas, Arūnas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD2154547&prefLang=en_US