%0 Journal Article
%T Composite Bayesian inference
%A Alexis Roche
%J Statistics
%D 2015
%I arXiv
%X This paper revisits the concept of composite likelihood from the perspective of probabilistic inference, and proposes a generalization called "super composite likelihood" for more flexible use of data information. It is shown that super composite likelihood yields a class of discriminative models suitable for unsupervised or weakly supervised learning.
%U http://arxiv.org/abs/1512.07678v1