Date

2013

Author

Özay, Mete

Metadata

Show full item record

Item Usage Stats

174
views

110
downloads

Cite This

In this thesis, Decision Fusion approaches have been analyzed for Supervised, Unsupervised and Semi-supervised Learning problems. In Supervised Learning, classification or generalization error minimization problem has been studied by analyzing the classification error of a classification algorithm into two parts. In the first part, the minimization of the difference between N-sample and large-sample classification error of k-NN has been studied using a hierarchical decision fusion algorithm called Fuzzy Stacked Generalization (FSG). Then, a weighted decision fusion and two sample selection algorithms are proposed to minimize the difference between large-sample error and Bayes Error in FSG. Unsupervised image segmentation problem has been analyzed for the fusion of decisions of different segmentation algorithms. An unsupervised decision fusion algorithm called Segmentation Fusion (SF) is proposed together with two distance learning methods. In addition, a weighted decision fusion method has been introduced. Two algorithms are suggested for the estimation of algorithm parameters and the number of different segmentation labels. The prior and side information about the statistical properties of data are integrated to SF using a new decision fusion algorithm called Semi-supervised Segmentation Fusion. The proposed algorithms and methods have been analyzed and examined on both synthetic and real-world datasets.

Subject Keywords

Data fusion (Statistics)., Statistical matching., Fuzzy algorithms., Computer algorithms., Segmentation fusion., Supervised learning (Machine learning).

URI

http://etd.lib.metu.edu.tr/upload/12615955/index.pdf
https://hdl.handle.net/11511/22798

Collections

Graduate School of Natural and Applied Sciences, Thesis