Abstract

Independent Component Analysis is becoming a popular exploratory method for analyzing complicated data such as that from functional magnetic resonance imaging (FMRI) experiments. The goal of ICA is to express a set of random variables as linear combinations of statistically independent component variables. This talk will describe the underlying generative model, different techniques for estimating the components, probabilistic extensions to ICA in order to overcome the “overfitting” problem, and possible generalisations to a higher-order decomposition for the joint analysis of multiple FMRI data sets. Examples of estimated artifactual component maps will be provided.