Learning with hidden variables
In many models some variables are not directly observed, but are latent or ‘hidden’. This can also occur when some data are not observed. In this chapter we discuss methods for learning in the presence of such missing information, and in particular develop the expectation maximisation algorithm and related variants.
11.1 Hidden variables and missing data
In practice data entries are often missing resulting in incomplete information to specify a likelihood. Observational variables may be split into visible (those for which we actually know the state) and missing (those whose states would nominally be known but are missing for a particular datapoint).
Another scenario in which not all variables in the model are ...