Semi-supervised Learning with Graphs

دانشگاه آزاد اسلامی اصفهان(خوراسگان)

منو

رکورد قبلی

رکورد بعدی

" Semi-supervised Learning with Graphs "

Document Type	:	Latin Dissertation
Language of Document	:	English
Record Number	:	150802
Doc. No	:	ET22594
Main Entry	:	Xiaojin Zhu
Title Proper	:	Semi-supervised Learning with Graphs
Note	:	This document is digital این مدرک بصورت الکترونیکی می باشد
Abstract	:	In traditional machine learning approaches to classification, one uses only a labeledset to train the classifier. Labeled instances however are often difficult, expensive,or time consuming to obtain, as they require the efforts of experienced humanannotators. Meanwhile unlabeled data may be relatively easy to collect, but therehas been few ways to use them. Semi-supervised learning addresses this problemby using large amount of unlabeled data, together with the labeled data, to buildbetter classifiers. Because semi-supervised learning requires less human effort andgives higher accuracy, it is of great interest both in theory and in practice.We present a series of novel semi-supervised learning approaches arising froma graph representation, where labeled and unlabeled instances are represented asvertices, and edges encode the similarity between instances. They address the fol-lowing questions: How to use unlabeled data? (label propagation); What is theprobabilistic interpretation? (Gaussian fields and harmonic functions); What ifwe can choose labeled data? (active learning); How to construct good graphs?(hyperparameter learning); How to work with kernel machines like SVM? (graphkernels); How to handle complex data like sequences? (kernel conditional ran-dom fields); How to handle scalability and induction? (harmonic mixtures). Anextensive literature review is included at the end.....-...,..tested for theQ1 PC1 bus cardBoth these projects mere sofixare des elopment efforts tonards contributing to dlfferentaspects of Roboucs and lZ1echatronics projects m the Controls and Roboucs Group..
Subject	:	Electericl tess
	:	برق
electronic file name	:	TL45818.pdf
Title and statement of responsibility and	:	Semi-supervised Learning with Graphs [Thesis]