Finite Mixture Models with Negative Components

Mixture models, especially mixtures of Gaussian, have been widely used due to their great flexibility and power. Non-Gaussian clusters can be approximated by several Gaussian components, however, it can not always acquire appropriate results. By cancelling the nonnegative constraint to mixture coefficients and introducing a new concept of negative components, we extend the traditional mixture models and enhance their performance without increasing the complexity obviously. Moreover, we propose a parameter estimation algorithm based on an iteration mechanism, which can effectively discover patterns of negative components. Experiments on some synthetic data testified the reasonableness of the proposed novel model and the effectiveness of the parameter estimation algorithm.