Gaussian Join Tree classifiers with applications to mass spectra classification
Víctor Bellón, Jesús Cerquides, Ivo Große · 2012
Classifiers based on probabilistic graphical models are very eective. In continuous domains, parameters for those classifiers are usually adjusted by maximum likelihood. When data is scarce, this can easily lead to overfitting. Nowadays, models are sought in domains where the number of data items is small and the number of variables is large. This is particularly true in the realm of bioinformatics. In this work we introduce Gaussian Join Trees (GJT) classifiers to try to partially overcome this issue by performing exact bayesian model averaging over the parameters. We use two die rent mass spectra classification datasets for cancer prediction to compare GJT classifiers with those learnt by maximum likelihood.