Extension of Mixture of Experts Model for Repeated Measures Data

• Activity Number: 247
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: IMS
• Author(s): Sungmin Myoung*+ and Chungmo Nam
• Companies: Jungwon University and Yonsei University College of Medicine
• Keywords: Mixture of Experts ; Linear Mixed Effect model ; Classification ; EM-algorithm

Abstract

The Mixture of experts (ME) is modular neural network architecture for supervised learning among a number of existed methods. This model can be considered as a mixture model that is consisted of the mixed distributions and weights in input variable x. Some of the researchers have been suggested to use of the multiple models for pattern classification and regression by ME method. However, the utilization of newly applied method is necessary in the repeated measures data. In this research, the mixture of experts is extended for repeated measures data. Moreover, cluster-specific effect is quantified via the linear mixed-effect model. To resolve this, firstly, we considered the construction of an expert, which has linear mixed-effect model, in ME. Afterward, the finding estimates were obtained for gating network and expert via EM-algorithm. The proposed model is more flexible than classical