Fast Transformation-Invariant Component Analysis

For software and more illustrations: http://www.psi.utoronto.ca/anitha/fastTCA.htm Dimensionality reduction techniques such as principal component analysis and factor analysis are used to discover a linear mapping between high dimensional data samples and points in a lower dimensional subspace. Previously, transformation-invariant component analysis (TCA) was introduced to learn this linear mapping in a way that is invariant to a set of global transformations. The expectation maximization algorithm used to learn the parameters of TCA requires a number of scalar operations on the order of N2 , where N is the number of elements in each training example. This is prohibitive for many applications of interest such as modelling mid- to large- size images, where N may be quite large (e.g. 262144 dimensions for a 512

• Type: Journal
• Year: 2008
• Where: IJCV
• Authors: Anitha Kannan, Nebojsa Jojic, Brendan J. Frey
• Keywords: Component Analysis, IJCV 2008, Linear Mapping, Lower Dimensional Subspace