Statistics Seminar

Title :INDEPENDENT COMPONENT ANALYSIS - THEORY, PRACTICE AND APPLICATIONS

Speaker: Dr Inge Koch
Department of Statistics
University of New South Wales

Time: 4:00p.m. Wednesday 3rd November 2004

Venue: Red Centre Building Room RC-3084
near Barker Street Gate 14

Independent Component Analysis (ICA) has been developed in the signal processing community as a powerful technique for 'blind source extraction'. Data x are given by x=As where s denotes the 'source' consisting of independent components, and A denotes the unknown transformation. Without distributional assumptions on s, ICA estimates the unknown A from observations x by optimising the non-Gaussianity of the independent components s. ICA has been applied to many and diverse problems
including neuroimaging, face recognition, speech processing, and predicting stock market prices.

In this talk we provide some background, explain the ICA methodology and practice, and show how ICA generalises classical multivariate methods such as PCA and Factor Analysis. We present some interesting standard applications in signal and image processing, and discuss recent advances. We conclude with some non-standard applications of ICA in areas as diverse as feature extraction and unsupervised learning.

Y. Fan
E-mail yanan@maths.unsw.edu.au
You are invited to join us for an afternoon tea in the staff common, Red Centre Building RC-3082 at 3:45pm.