Simple mixture model for sparse overcomplete ICA

M. Davies, N. Mitianoudis
2004 IEE Proceedings - Vision Image and Signal Processing  
We explore the use of Mixture of Gaussians (MoGs) for noisy and overcomplete ICA when the source distributions are very sparse. The sparsity model can often be justified if an appropriate transform, such as the Modified Discrete Cosine Transform, is used. Given the sparsity assumption we are able to introduce a number of simplifying approximations to the observation density that avoids the exponential growth of mixture components. We further derive an efficient clustering algorithm whose
more » ... orithm whose complexity grows linearly with the number of sources and show that it is capable of performing reasonable separation.
doi:10.1049/ip-vis:20040304 fatcat:zygodz3lxjdvbazzif5cmt2z5e