The label switching problem is caused by the likelihood of a Bayesian mixture model being invariant to permutations of the labels. The permutation can change multiple times betwee...
Abstract—It is well known that for finite-sized networks, onestep retrieval in the autoassociative Willshaw net is a suboptimal way to extract the information stored in the syna...
In many signal processing problems, it may be fruitful to represent the signal under study in a redundant linear decomposition called a frame. If a probabilistic approach is adopt...
Bayesian networks with mixtures of truncated exponentials (MTEs) support efficient inference algorithms and provide a flexible way of modeling hybrid domains (domains containing ...
Helge Langseth, Thomas D. Nielsen, Rafael Rum&iacu...
The practical success of broadcast encryption hinges on the ability to (1) revoke the access of compromised keys and (2) determine which keys have been compromised. In this work w...