================== Observation Models ================== All observation models define a *likelihood* for producing data :math:`x_n` from some cluster-specific density with parameter :math:`\phi_k`: .. math :: p(x | \phi, z) = \prod_{n=1}^N p( x_n | \phi_k )^{\delta_k(z_{n})} Supported Bayesian methods require specifying a (conjugate) prior: .. math :: p(\phi) = \prod_{k=1}^K p(\phi_k) Variational methods for observation models ------------------------------------------ The links below describe the mathematical and computational details for performing standard variational optimization for supported observation models: .. toctree:: :maxdepth: 1 ZeroMeanGaussObsModel-VB DiagGaussObsModel-VB GaussObsModel-VB GaussRegressYFromFixedXObsModel-VB