creators_name: Valpola, Harri
type: techreport
datestamp: 2004-05-14
lastmod: 2011-03-11 08:55:36
metadata_visibility: show
title: Behaviourally meaningful representations from normalisation and context-guided denoising
ispublished: pub
subjects: comp-neuro-sci
subjects: comp-sci-mach-learn
subjects: comp-sci-neural-nets
subjects: comp-sci-art-intel
full_text_status: public
abstract: Many existing independent component analysis algorithms include a preprocessing stage where the inputs are sphered.  This amounts to normalising the data such that all correlations between the variables are removed.  In this work, I show that sphering allows very weak contextual modulation to steer the development of meaningful features. Context-biased competition has been proposed as a model of covert attention and I propose that sphering-like normalisation also allows weaker top-down bias to guide attention.

date: 2004-05
date_type: published
institution: University of Zurich
department: Artificial Intelligence Laboratory
refereed: FALSE
citation:   Valpola, Harri  (2004) Behaviourally meaningful representations from normalisation and context-guided denoising.  [Departmental Technical Report]     
document_url: http://cogprints.org/3633/1/tr04a.pdf