Discrete Component Analysis
Dr Wray Buntine, Senior Research Scientist (Helsinki Institute for Information Technology (HIIT))
NICTA SML SEMINARDATE: 2006-07-21
TIME: 13:00:00 - 14:00:00
LOCATION: RSISE Seminar Room, ground floor, building 115, cnr. North and Daley Roads, ANU
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
The first part of this talk will review the model and discuss some alternatives. The model is a discrete variant of PCA (principal components analysis) and ICA (independent components analysis) and is an instance of latent variable modelling. It provides a useful starting point for more detailed models now used in text analysis and genetics. Wellknown variants are non-negative matrix factorisation and latent Dirichlet allocation. Some variants presented will be sparse versions (making the loading matrices and score matrices sparse), semi-supervised versions, and dealing with n-grams in the content.
Basic algorithms will be reviewed (but not presented in detail) and some extensions to the basic model will be discussed. Experimental evidence will also be presented for some of the methods, since the algorithms make an excellent case study in alternative methods for statistical computing.
See http://www.componentanalysis.org for the software.
