General learning algorithms based on finite mixture density models have recently been developed and applied to independent component analysis and the blind source separation of linear mixtures. This paper proposes the finite mixture density originally proposed by Pearson as a simple parametric model to be used within these algorithms for clustering and visualising data. The various forms of the resulting algorithms, which utilise this mixture model, bring fresh insights to the nature of the robust principal component analysis and extended infomax independent component analysis algorithms.
展开▼