Mixture model is known as a convenient way for modelling the probability density function in statistics. Recently, the method is adopted by machine learning communities in a variety of application settings such as cluster analysis, classification, density estimation and function approximation. This book concerns with learning algorithms of the mixture models for density estimation and classification tasks. Special attention is given for the semi-supervised learning and active learning methods which are very important in many practical settings. The presented learning methods attempt to reduce the size of labelled data sets required to achieve certain level of classification performance.