The FMM procedure makes the following assumptions in fitting statistical models:

The number of components k in the finite mixture is known a priori and is not a parameter to be estimated.

The parameters of the components are distinct a priori.

The observations are uncorrelated.