Fitting large mixture models using stochastic component selection

Milan Papež, Tomáš Pevný, Václav Šmídl

Traditional methods for unsupervised learning of finite mixture models require to evaluate the likelihood of all components of the mixture. This becomes computationally prohibitive…