Publication

Abstract : The present and possible future roles of Mixture of Experts and Hierarchical Mixture of Experts models, especially for pattern recognition is discussed and reviewed. Both supervised and unsupervised techniques are explored along with the various techniques for parameter estimation. The EM algorithm, which forms a vital part in most of the parameter estimation algorithms, is discussed in detail and its various forms of maximization procedures. A synthesis is made of the unresolved problems related to its application in image classification. Finally an outlook into the future applications of ME and HME has been presented.