Loading...

Regularization from the Machine Learning Point of View

Ghaemi, Mohammad Sajjad | 2010

533 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 41112 (02)
  4. University: Sharif University of Technology
  5. Department: Mathematical Sciences
  6. Advisor(s): Daneshgar, Amir
  7. Abstract:
  8. In traditional machine learning approaches to classification, one uses only a labeled set to train the classifier. Labeled instances however are often difficult, expensive, or time consuming to obtain, as they require the efforts of experienced human annotators. Meanwhile unlabeled data may be relatively easy to collect, but there has been few ways to use them. Semi-supervised learning addresses this problem by using large amount of unlabeled data, together with the labeled data, to build better classifiers. Because semi-supervised learning requires less human effort and gives higher accuracy.Formally, this intuition corresponds to estimating a label function f on the graph so that it satisfies two things:The prediction f(x) is close to the given label y on labeled vertices. f Should be smooth on the whole graph.This can be expressed in a regularization framework.In this thesis, several different semi-supervised learning algorithms based on graph will be introduced based on the following papers. They differ in the choice of the loss function and the regularizer.[Abney, 2008, Belkin et al., 2006, Bengio et al., 2006, Blum and Chawla, 2001, Chapelle et al., 2003, Dai and yan Yeung, 2007, Dhillon et al., 2004, Duda et al., 2001, Hagen and Kahng, 1992, Kondor and Lafferty, 2002, Kulis et al., 2005, Lebanon, 2005, Shi and Malik, 2000, Smola and Kondor, 2003, Steinwart and Christmann, 2008, Theodoridis and Koutroumbas, 2008, von Luxburg, 2006, Zhou et al., 2004, Zhu and Goldberg, 2009, Zhu et al., 2003, Zhu et al., 2004]
  9. Keywords:
  10. Unsupervised Learning ; Semi-Supervised Learning ; Regularization ; Kernel Matrix ; Supervised Learning

 Digital Object List

 Bookmark

No TOC