Creating a DPM

The first step is to define a subtype of AbstractDPM that extends the type DPM using composition, e.g.

struct MyDPM <: AbstractDPM
    # ... any desired fields

In this way, all the standard components of the DPM (e.g. the vector of cluster labels) will be stored in parent. Please, read carefully the documentation about DPM and its constructor/accessors in order to avoid redundancies.

Once MyDPM is defined, the second and last step is to specialize the following methods to m::MyDPM:

update_suffstats!(m::MyDPM, data)
update_suffstats!(m::MyDPM, data, i::Int, k0::Int, k1::Int)
update_hyperpars!(rng::AbstractRNG, m::MyDPM, data)
logpredlik(m::MyDPM, data, i::Int, k::Int)

see the documentation of each method for more details. Once again, we note that data can have any type, as long as each method is correctly implemented.