Bayesian feature weighting for unsupervised learning, with application to object recognition

Peter Carbonetto (1) , Nando de Freitas (1) , Paul Gustafson (2) , Natalie Thompson (2)

1 UBC-Computer Science - Computer Science Department
2 UBC Statistics - Department of Statistics [Vancouver]

We present a method for variable selection/weighting in an unsupervised learning context using Bayesian shrinkage. The basis for the model parameters and cluster assignments can be computed simultaneous using an efficient EM algorithm. Applying our Bayesian shrinkage model to a complex problem in object recognition (Duygulu, Barnard, de Freitas and Forsyth 2002), our experiments yied good results.

Mots clés: LEAR

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Peter Carbonetto, Nando de Freitas, Paul Gustafson, Natalie Thompson. Bayesian feature weighting for unsupervised learning, with application to object recognition. Artificial Intelligence and Statistics (AI & Statistics '03), Microsoft, Jan 2003, Key West, United States. ⟨inria-00548235⟩