Multitask learning in mixture modelling framework via generalized linear mixed-effects models
MetadataShow full item record
Many real-world problems can be considered as a series of related tasks. For example, related tasks are to predict survival of patients from di erent hospitals. In these multitask problems, the data collected could exhibit a clustered structure due to the relatedness between multiple tasks. Mixture model-based methods assuming independence may not be valid for regression and cluster analyses of data arisen from multiple related tasks. Multitask learning is an inductive transfer mechanism to improve generalization accuracy by sharing task-speci c information from di erent tasks to improve the learning process. In this paper, the multitask learning mechanism is extended for mixtures of generalized linear models via random-e ects modelling to handle multitask problems. The use of random-e ects models implies that a soft sharing mechanism is adopted to leverage task-speci c information from multiple tasks. The proposed method is illustrated using simulated and real data sets from various scienti c elds.
20th International Conference on Computational Statistics
Copyright 2012 International Association for Statistical Computing. The attached file is reproduced here in accordance with the copyright policy of the publisher. Please refer to the conference's website for access to the definitive, published version.
Medical and Health Sciences not elsewhere classified