|
hi i have some observations at high dimension (hyperspectral imagery). and i want to unmix the components in the observed data while i know that some of the components are dependent to each other (mineral exploration application of hyperspectral remote sensing)! can i ask for solution of demixing of dependent component only based on observations? is gibbs sampler associated with Dirichlet mixture models useful for this task? please guide me for this purpose. thanks a lot.
This question is marked "community wiki".
|
Dirichlet Mixture Models work for doing clustering of objects (in a sense). If you have dependency in your clusters, you may use hierarchical mixture models, where you model those dependencies. Gibbs sampler is nothing both one (of many) ways to obtain the probability sample associated with the model.